Answers to Selected Exercises

Worked solutions and model critiques for the starred and odd-numbered exercises from each chapter. Video is learned by shooting and cutting — attempt every assignment with a camera or an edit timeline in hand before reading its solution.

Chapter 1 — Answers to Selected Exercises

1 (The four-second test). Strong answers name a specific cause, not a vibe: "a face already mid-sentence with a bold claim," "motion toward the camera," "a question I wanted answered," or, for a loss, "five seconds of a logo animation," "a slow, silent pan with nothing happening," "muffled audio." The lesson: attention is won or lost by concrete choices in the first seconds — a preview of Chapter 23's "hook."

3 (Name the stages). Any reasonable reconstruction works; the point is recognizing that decisions live in all three stages. Example for a vlog: pre — deciding the video's one topic and where to film for good light; production — clipping on a mic and shooting a few takes plus B-roll; post — cutting the rambling down, adding music and captions, exporting. If a student can only find "production," push them: someone chose the topic (pre) and cut it (post).

5 (The reverse-engineer). A full seven-field Described Shot; grade on precision, not correctness. Good answers commit to specifics ("medium close-up, subject on the right third; slow push-in; soft key from window left; music bed with no dialogue; hard cut in on the first word; the push-in pulls us toward the emotion; lesson: a slow push-in intensifies a moment"). Weak answers stay vague ("nice shot, good lighting"). The skill being built is specificity.

7 (Diagnose the field — the push-in). With a slow push toward the door as the person enters, THE MOVE becomes active and THE EFFECT shifts from "observe the calm place" to "we are drawn toward this person; something is beginning." The move adds anticipation the locked-off version withholds. Neither is "right" — they serve different intentions. The lesson: the move is a storytelling choice, not a default.

8 (The missing piece — FIGURE 1.3). Many valid answers. E.g., add a shot of an empty chair across the table and cut it after the neutral face: now the smile can read as "seeing someone arrive." Or add a shot of a bill/check: the smile reads differently again. The lesson (reinforcing the Kuleshov effect): each added shot recontextualizes the ones around it. Meaning is assembled.

9 (Write your own Described Sequence). Grade on whether the table captures shot sizes, durations, audio, and why it cuts where it cuts. The paragraph should explain a pattern (e.g., "it cuts on movement so the edits feel invisible" or "it holds the reaction longer than the action"). This is deliberately hard this early; effort and specificity are the win.

10 (Twenty deliberate seconds). A model self-critique names one concrete strength and one concrete, fixable flaw: "Take 2 is best because the light was on my face and the framing was steady; but the audio has fridge hum I should have killed, and I cut off too fast — I needed a beat of room tone at the end." Self-critique that names fixable causes (not "it was bad") is the target skill.

12 (Five Ways). No single right answer; the value is discovering that shot size and angle change meaning, not just look. Typically the close-up or the low angle "tells the story of the action" most clearly because it isolates and emphasizes. Students should be able to say why their pick worked — that reasoning is the learning.

15 (Fix the plan). Three skipped decisions, e.g.: (1) light — sitting at a desk with no thought to where the light falls risks a dim or unflattering face; face a window. (2) sound — a phone across the desk in an untreated room will sound hollow; get a mic close and reduce room noise. (3) story/length — "talk for a few minutes" has no structure; decide the single message and the first-sentence hook. Bonus: framing/background, and a second take.

16 (Fix the shoot — the backlit interview). Problems and fixes: backlight/silhouette → turn the subject toward the window or add light on the face (pre-production: plan where they face; production: reposition). Echoey audio in a bare room → get the mic close and add soft furnishings, or move to a smaller/softer room (pre: scout a better room; production: mic placement). The meta-lesson: most of these are cheapest to prevent in pre-production.

17 (The story is the boss — bakery). A strong sentence is specific and human: e.g., "Every loaf here is shaped by hand at 4 a.m. by a baker who learned from her mother." Shots: hands shaping dough (detail), the dark early bakery (wide/mood), the baker's face (medium), the finished loaves (detail), a customer's first bite (payoff). A weak answer ("we sell fresh bread") produces generic shots; the specific angle produces a video.

18 (The impossible rescue). Three things the edit cannot fix if not captured: (1) a reaction/cutaway that was never shot — you cannot cut to what does not exist; (2) clipped or distorted audio — once the sound is destroyed at capture, no plugin restores it cleanly; (3) a missing shot size/angle — you cannot invent a wide (or a close-up) you never filmed, so you cannot shorten or reshape the scene. All three trace to one principle: post arranges; it does not create. Shoot for the edit.

20 (Define Project 1). A strong definition names a real person, one message, and a compelling reason to watch: "My grandfather, telling the one piece of advice he wishes he'd taken at 20 — because everyone recognizes the regret." A weak one is vague: "Me talking about my week." Push students toward a single idea and an actual hook.

23 (Assemble your ritual). Guidance: the win is realizing that selection and order are decisions, not clerical steps. A good assembly picks the best take (steady, in focus, well-lit), orders it wide→medium→detail for legibility, and trims dead air off each clip's ends. Students should notice that a different order produces a different feel — that noticing is the point.

25 (Two cuts, one message). Grade on whether the two edits genuinely read differently to a fresh viewer. "Peaceful" might use longer holds, slower moments, and calmer shots; "lonely" might use wider framing, emptier shots, and a slower, sparser rhythm — from the same footage. The lesson: meaning is manufactured in the edit through selection, order, and pace.

26 (Read the box). The six: orientation (matches where it will be shown), resolution/frame rate (quality and motion feel), stability (steady reads as competent), light (where it falls decides the whole look), sound (half the picture), lens (the main camera is sharpest). Full explanations are in Chapters 2–5 and Part III.

27 (Write the settings). Model instincts: (a) horizontal, subject facing a window, mic close, single clear message — the classic talking-head. (b) Vertical (9:16) for a phone feed, side light on the food, close sound, cut to the rhythm of the action. (c) The café: the hard problem is sound — get a mic as close as possible, pick the quietest corner and time, and expect to fight room noise (Chapter 15 has the real tools). All three are first instincts to be refined later; reward sound reasoning over "correct" answers.

28 (Teach it back). A strong ~200-word explanation uses the three stages, at least one craft element, and a concrete example to show that a video is built from decisions, not captured by a device. Grade on clarity for a true beginner and on whether the example actually illustrates the point (e.g., the café-owner testimonial from §1.2).

29 (The self-audit). No right answer; the value is an honest, specific diagnosis. Strong self-audits name a concrete, fixable weakness ("the audio was echoey; I'll get the mic closer") rather than a vague one ("it wasn't good"). Whichever element scored lowest points to where the book will help most next.

Chapter 2 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises (numbering matches the 40-exercise set in exercises.md). Grade for reasoning tied to the image and the story, not for a single "correct" number — most of these have a range of good answers.

1 (Spot the frame rate). A strong answer names the motion difference, not a vibe: the film's motion has a subtle judder/weight that reads as "cinematic," while the sports/soap footage is glassy-smooth and "live." Lesson: 24 fps vs 60 fps is a look, visible to the trained eye once you know to watch for it.

2 (The soap-opera hunt). The culprit is almost always a high frame rate (60 fps) — or a TV's motion-smoothing/interpolation. Accept any answer that identifies smoothness-of-motion as the "cheap" tell rather than resolution or sharpness. Reinforces that "cheap-looking" is often a frame-rate problem, not a gear problem.

3 (Motion-blur watch). On a filmic frame the moving thing is blurred (a soft streak); on a video-game or high-shutter clip it's frozen sharp. Good answers connect the blur to how "filmic" the motion felt in playback: natural blur → smooth/cinematic; frozen frames → stuttery. This is the §2.4 mechanism observed with the film paused.

4 (Find the rolling shutter). Not graded for a "right" answer — the win is simply spotting real skew (leaning poles from a car window, a wobbling pan, a rubbery guitar string) and describing it. Once seen, it can't be unseen, which is the point.

5 (The resolution can't-tell test). The honest answer for most students, at normal distance on a normal screen, is "I can barely tell them apart." The lesson isn't that resolution is useless — it's that its main value is reframe headroom (what you do in the edit), not visible delivery sharpness. A student who says "I could tell on a huge TV up close" is also right; the takeaway is to match resolution to the use, not to maximize it reflexively.

6 (Name the setting decision). The four choices in FIGURE 2.8: 4K (so the edit can punch in / reframe), 24 fps (filmic feel for a testimonial), 1/50 shutter (natural motion via the 180° rule), locked off (steadiness says "listen," keeps reframe headroom clean). Full marks require the reason with each, not just the setting.

7 (Diagnose the field — flip the frame rate). At 60 fps, THE EFFECT becomes hyper-smooth and "live" — present and immediate, but the piece loses its filmic gravity and feels like a video call or news hit. THE LESSON: frame rate changes the emotional register even when nothing else changes. A beginner thinks 60 is "better" because the number is higher and the motion is smoother — mistaking smoothness for quality, the exact trap §2.3 warns about.

8 (Read the walk two ways). 60 fps wins when the story wants immediacy, energy, realness, or a planned slow-motion moment — a sports piece, an action reel, a beauty slow-mo. 24 fps wins when the story wants film feel, gravity, narrative distance — a drama, a brand story, a doc. The discriminator: does this moment want to feel like cinema or like live/real?

9 (Write your own settings-Described-Shot). Grade on whether THE FRAME/MOVE/EFFECT name real settings and tie each to the story. Strong: "4K so I can punch in on the key line; 24 fps because this confession should feel like a film; 1/50 for natural motion; locked-off so the stillness says 'listen.'" Weak: settings listed with no story reason. Reward specificity and motivation.

10 (Critique the case-study homage). Model: a reduced shutter exposes each frame for a much shorter slice of time, so moving subjects are caught nearly frozen with little motion blur; on playback the eye, which expects natural blur, reads the sharp, discontinuous frames as stuttery and "wrong." That wrongness suits chaos. An everyday use: a frantic action montage, a tense chase, a deliberately unsettling moment — anywhere jarring motion serves the feeling. Warn against making it a default.

11 (The three-shutter test). A model critique names the feel of each: 1/50 = natural, invisible, correct; 1/500 = stuttery, robotic, "gatey," subtly cheap; 1/24 = smeary, dreamlike, wrong for anything crisp. The win is recognizing that the correct shutter is the one you stop noticing, and being able to identify the two failure modes by eye.

12 (The frame-rate A/B). Expect: 24 fps feels like a film, 60 fps feels like a phone/live clip, same everything else. Best answers also note the side effect: keeping the correct 180° shutter meant it changed between takes (1/50 → 1/120), which changed how much light hit the sensor — a first felt encounter with the exposure trade-off of Chapter 5.

13 (Slow-motion for real). Any sensible placement earns full marks; the point is recognizing where a planned slow beat adds value — a wedding first-look, a product hero shot, a sports highlight, an emotional pause. The deeper insight: the smoothness comes from real captured frames, which is why you decide slow-motion on set (by choosing the high frame rate), not in the edit.

14 (Provoke rolling shutter). Not graded for correctness. The rule it should teach, in the student's own words: for anything that moves, slow the camera and stabilize — fast whip-moves on a rolling-shutter sensor skew straight lines.

15 (The reframe-headroom proof). Guidance: success = two usable "shots" (a wide and a punched-in medium) cut from one 4K take, with the punch-in still clean at 1080. The realization: you created coverage after the shoot because you captured enough resolution to afford it — the literal meaning of "you shoot for the edit." Watch for students pushing the crop past a 1080 window and softening; that's the edge of reframe headroom.

16 (Sensor-size, felt). With two cameras: the bigger sensor typically shows a blurrier background at matched framing and cleaner low light. With one camera: the image degrades (noisier/grainier) as light drops, because the photosites are starved (§2.1). Model lesson: sensor size is a trade, and light is usually the bigger lever — which is why Part III matters so much.

17 (The four-setup settings sweep). Model: in the talking-head (still), resolution/reframe headroom matters most — the camera doesn't move, so rolling shutter and even shutter are low-stakes; you fuss over framing and whether to shoot 4K for punch-ins. In the walk-and-talk (moving), shutter and rolling shutter dominate — a wrong shutter stutters every step, and fast moves skew. Lesson: the setup decides which of the four settings you sweat. Same camera, different worries.

18 (Five frame rates / shutters of one action). No single right winner; reward the defense. Typically (a) 24/1-50 reads natural-filmic; (b) 24/1-500 reads tense/staccato; (c) 60 normal reads smooth/live; (d) 60→24 gives clean slow-motion; (e) high slow-mo is dreamy/dramatic. The learning is that motion-related settings change meaning, and the best choice depends on the feeling the action wants.

19 (Five resolutions / crops). Model finding: 720p/1080p/4K look nearly identical on a normal screen; the 4K-with-a-1080-crop stays clean until you crop past a 1080 window, at which point it softens. Lesson: reframe headroom is generous but finite — you can punch in roughly to a 1080 window inside a 4K frame before quality visibly drops.

20 (Write the shutter). 24 fps → 1/50 (1/48 rounded); 25 fps → 1/50; 30 fps → 1/60; 60 fps → 1/120 (or 1/125). Rule: shutter ≈ 1 ÷ (2 × frame rate).

21 (Four briefs, four baselines). Models: (a) filmic testimonial — 4K (reframe) or 1080, 24 fps, 1/50; filmic feel. (b) sports highlight for social — 1080 or 4K, 60 fps (smooth action + slow-mo option), 1/120; likely vertical (Ch.23). (c) wedding with dreamy slow-mo — 4K, 24 fps main coverage plus 60/120 fps inserts for the slow beats, 180° shutter for each; label the slow-mo shots. (d) heavily-reframed locked-off product — 4K (max reframe headroom), 24 or 30 fps, 1/50 or 1/60; locked off. Grade the reasoning, not the numbers.

22 (The bright-day problem). Three motion-safe fixes: add a neutral-density (ND) filter; stop down the aperture and/or lower the ISO (Chapter 5); or move to shade / wait for softer light. Must NOT: raise the shutter (it freezes motion). Chapter 5 solves this fully.

23 (The unfamiliar camera). Model checklist: (1) set resolution + frame rate for the project (e.g., 4K, 24 fps); (2) set the shutter to the 180° value (1/50); (3) set exposure (ISO/aperture, or ND) without touching the shutter; (4) set white balance and confirm audio is recording. If you can't find the manual shutter: turn off full auto if you can, or shoot in soft even light where the auto shutter behaves and avoid harsh light. Bonus: shoot a 10-second test and watch it back.

24 (The social spec). Lean 30 or 60 fps here (not 24): fast-cut, punchy social content often benefits from smoother, more energetic motion, and 60 fps also gives you slow-mo options and holds up under heavy re-encoding better on some platforms; the "live/immediate" feel that hurts a filmic testimonial actually suits the format. Frame vertical (9:16). Full treatment in Chapter 23. Accept 24 fps if defended as a deliberate "filmic on social" style — the point is the reasoning.

25 (Fix the stutter). Most likely the shutter is too high (the camera on auto pushed it to ~1/1000 in bright light), freezing motion into a stutter. Fix: set the shutter to the 180° value for the frame rate (1/50 at 24 fps) and control brightness with ND/ISO/aperture instead.

26 (Fix the cheap-looking brand film). The culprit is almost certainly a high frame rate (60 fps) giving the smooth "soap opera"/news look. On the reshoot, shoot 24 fps (25 in PAL) at a 1/50 shutter. Nothing about focus or exposure needed changing — it was the motion.

27 (Fix the melting building). The effect is rolling shutter skew: the sensor reads line by line, so during fast drone moves the bottom of the frame lags the top and vertical structures lean/wobble. Two fixes: slow the moves and stabilize; a global-shutter camera avoids it entirely.

28 (Fix the impossible edit). On a 24 fps timeline the 60 fps clip plays slowed (its extra frames stretch over more time) unless told to play normal speed, while the 24 fps clip plays normally — so intercutting them "at normal speed" makes the 60 fps material behave unexpectedly, and conforming introduces stutter or frame-blending. The shooter should have picked one frame rate for the project (or shot the 60 fps material as planned slow-motion and labeled it). Meta-lesson: frame-rate discipline on set prevents edit-day pain.

29 (Fix the storage disaster). Two mistakes: (1) shooting 8K to deliver 1080 — vastly more resolution than the job needs, for no visible benefit; and (2) editing the huge native files directly on an underpowered laptop. Fixes: shoot a sane resolution for the job (1080, or 4K if reframe is wanted), and edit with proxies — lightweight stand-in copies that keep the timeline fast, swapped for the originals on export. Proxies are Chapter 27. Resolution is a trade, never a free "safety."

30 (The reduced-shutter look). Guidance: the high-shutter version feels tense and "wrong" because stripping the motion blur makes movement sharp and staccato — the eye recoils from motion it expects to be blurred (the §2.4 mechanism and the Omaha Beach look of Case Study 1). Two sentences should connect the look to a feeling (chaos, urgency) and acknowledge it's borrowed on purpose, not a default.

31 (The dreamy smear). At 24 fps with the widest shutter (near 1/24 / 360°), quick motion smears into soft streaks — useful for dream sequences, memory, intoxication, a deliberate woozy transition. It looks like a mistake on anything meant to be crisp. Lesson: like the reduced shutter, the wide shutter is a motivated departure, not a default.

32 (The slow-motion beauty shot). Real high-frame-rate capture looks better than software-slowed 24 fps footage because it contains real captured frames to fill the stretched time; slowing 24 fps footage in software forces the computer to invent in-between frames (or repeat/blend existing ones), which looks stuttery or smeary. The takeaway: decide slow-motion on set by choosing the high frame rate — you can't add real frames later.

33 (Conform a mismatch). On a 24 fps timeline: the 60 fps clip plays in slow motion by default (more captured frames stretched over the timeline's time), and the 24 fps clip plays normally. One-line takeaway: matching frame rates on set means clips "just fit" the timeline — no unexpected slow-mo, no stutter, no conform guesswork.

34 (Build the two-angle cut). Guidance: success = a 20-second edit cutting at least twice between the wide and the 1080 punch-in of the same 4K take, timed to the subject's words (cut to the punch-in on the emphasis line, back to the wide to breathe). The realization: you edited coverage that never existed as separate shots — it was reframe headroom. If the crop softened, they pushed past ~1080; that's the practical limit.

35 (Mix a slow-mo insert correctly). On the 24 fps timeline, the 24 fps main shot plays normally and the 60 fps insert, interpreted at its captured frame count, plays smooth-slow — exactly what you want, because you planned the 60 fps clip to be slow. This is the right way to mix rates: one project frame rate (24) plus deliberate, labeled high-frame-rate inserts for slow-motion. Exercise 28's mismatch was accidental — the same mechanism, but unplanned, so it fought the edit. Intent is the whole difference.

36 (The delivery-rate check). Exporting a 24 fps piece at 30 fps (or vice versa) forces the software to add/drop/blend frames to fit, introducing subtle judder or stutter, especially on motion and pans. The rule it proves: deliver at the frame rate you shot (and edited) unless you have a specific reason and the tools to convert cleanly.

37 (Settings serve story). Model: "My Project 1 is my grandmother telling the one recipe she never wrote down — I'm shooting 24 fps because this should feel like a keepsake film, not a video call; 1/50 for natural motion; 4K so I can punch in on her hands and her face from one take." Weak answers can't connect the number to a feeling — flag those; arbitrary settings are the thing this exercise exists to cure.

38 (The three-stages check). A camera test is a pre-production/prep step (you're de-risking before the real shoot). A free catch it provides: discovering that the camera was defaulting to a high auto-shutter (stuttery motion) or the wrong frame rate — either of which is trivial to fix on a throwaway test clip but impossible to fix in the edit if it's baked into a performance you can't re-shoot. Directly embodies "fix it in pre, not in post."

39 (Teach the four settings back). A strong ~200-word explanation shows that even though this chapter is "about the camera," the settings only matter as decisions in service of the image: choosing 24 fps isn't a technical act, it's deciding the video should feel like a film; choosing 4K isn't "more," it's giving the future edit room to reframe. The point to land: the numbers are how you express a decision the story already made — so "the video is not the footage" still holds. Grade on clarity for a true beginner and on using at least two settings to prove the point.

40 (The all-three-projects settings page). Model: talking-head — 4K/24/1-50 (intimate, filmic, reframe headroom); documentary short — 4K/24/1-50 for interviews and B-roll, with labeled 60 fps inserts for any slow beats (a film feel with flexibility); branded piece — 4K/24/1-50, possibly a 60 fps hero slow-mo for the product, plus a vertical 30/60 fps social cut-down (Ch.23). Reward students who justify differences from the piece's feeling rather than copying one baseline everywhere — and who recognize most storytelling lands on 24/1-50 with slow-mo as the main exception.

Chapter 3 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises (numbering matches the 40-exercise set in exercises.md). Grade for reasoning tied to the post-production consequence, not for a single "correct" format — most of these have a range of good answers, and the point is always "what does this choice cost or buy in the edit and grade?"

1 (Spot the banding). Not graded for a "right" answer — the win is simply spotting real banding (a sunset stepping into stripes, a studio backdrop breaking behind a host, a fade to black in bands) and describing it. Most common on heavily-compressed streams and social re-uploads, where bitrate and bit depth are both squeezed. Once seen, it can't be unseen.

2 (Hunt for over-compression). Accept any answer that correctly names the artifact (macroblocking = blocky squares; smearing = mushy motion; banding = stepped gradients) and connects it to low bitrate on a hard-to-compress scene (shadows, motion, gradients). Reinforces that "cheap-looking internet video" is usually a bitrate problem.

3 (Re-upload degradation). Expect: gradients band, shadows block up, fine motion smears — the platform re-compressed the file at a lower bitrate. The lesson: lossy compression is cumulative, and every re-encode throws away more. This is why you keep a high-quality master and deliver from it (Ch.36).

4 (Standard vs graded). A strong answer reads the color as built or captured: a heavy teal-orange split, a warm nostalgic wash, or crushed bleached blacks almost always signal a grade, not a straight-from-camera image. The tell is usually that the color is too consistent, too extreme, or too motivated to be an accident. Sets up Case Study 1.

5 (Dynamic-range audit). Not graded for correctness — the win is noticing the choice: did the shot hold detail in both the window and the face, or sacrifice one (blown window / black face)? Good answers name which end each example protected. This trains the eye for §3.4 and Chapters 5/11/13.

6 (Name the format decision). The standard profile gave up latitude (it clipped highlights and crushed shadows at capture to look good immediately). Log gave up watchability (it looks flat/grey/wrong out of camera, and it requires a grade, bit depth, and careful exposure) to keep maximum range and color for post. Full marks require naming both sacrifices.

7 (Read the ladder). Grade the defense, not the placement. Typical: standard 8-bit 4:2:0 → a fast talking-head or social clip (light color, tight turnaround); Log 10-bit 4:2:2 → a branded piece or music video with a real grade; RAW → a high-end commercial or VFX shot with the pipeline to carry it. Reward a one-sentence reason tied to the post plan.

8 (Diagnose the pipeline figure). Renaming .MP4 to .MOV only changes the container label; the codec inside is unchanged, so if the software couldn't decode that codec, it still can't. What actually fixes it: transcode the file into a friendly codec (or install the decoder) — a contents fix, not a label fix.

9 (Write a format Described Shot). Grade on whether THE FRAME/EFFECT name real capture settings and tie each to post. Strong: "10-bit Log at high bitrate because I'll build a warm grade; exposed slightly bright to keep the shadow behind her clean; a standard clip would band when I push it." Weak: settings with no post-reason. Reward specificity and motivation.

10 (The café grade, previewed). To make the future warm café grade possible you'd want: enough dynamic range (a flat/Log or wide-range profile) to hold the window and the interior; enough bit depth (10-bit) so warming the whites and deepening the amber doesn't band; a healthy bitrate. Shot 8-bit 4:2:0 with a contrasty look baked in, the heavy warm grade would band the walls, block the shadows, and have little room to move the color — the look would be largely impossible.

11 (The banding test). A model critique names where and how fast it bands: a smooth wall or dusk sky steps into visible stripes under a hard contrast/shadow/tint push, and it worsens as you push. The win is recognizing that 8-bit heavily-compressed footage has a low ceiling, and (with a 10-bit clip) seeing the bands come far slower or not at all. Keep both clips as proof.

12 (Bitrate A/B). Expect: on a busy, hard-to-compress scene (foliage, water, crowd), the low-bitrate clip smears and blocks where the high-bitrate one holds detail. Best answers note that scene difficulty decides how much bitrate you need — a locked-off wall survives a low bitrate; wind in grass doesn't.

13 (Standard vs Log, graded). Guidance: success = the Log clip, once brought back to normal (color-space transform or LUT) and graded, gives more control and holds the push better than the standard clip — while looking worse than the standard clip before grading. That gap (worst ungraded, best graded) is the entire standard-vs-Log decision, seen firsthand.

14 (White-balance rescue). Expect: the standard clip fights the correction (color casts, degraded skin) because white balance was baked in; the RAW clip corrects cleanly because white balance wasn't committed at capture. Lesson: RAW makes white balance a near-free post decision — a genuine safety net — because it records sensor data before the profile.

15 (Green-screen edge test). Expect: the 8-bit 4:2:0 version keys with soft, blocky edges; the 10-bit (or higher-chroma) version keys cleaner. The connection: keying needs clean color edges, and 4:2:0's quarter-resolution color gives the matte poor edges. More chroma (4:2:2) and more bit depth = a cleaner key. (Green screen: Ch.35.)

16 (Shoot the dynamic-range problem). Grade the paragraph: which end they protected (face or window) and why, and the recognition that no single exposure holds both because the scene exceeds the camera's range. The tools that let you keep both: light/bounce the shadow side, move the subject, or shoot flatter/Log to hold more range — previewing Chapters 5, 11, 13.

17 (Four-setup format sweep). Model: talking-head → standard high-bitrate (light correction); walk-and-talk → Log/HLG high-bitrate (fighting sky-vs-face dynamic range); product → 10-bit (color-critical, smooth backdrops that band); event → efficient codec, reliable bitrate, fast offload (run-and-gun, long records). The format should change between setups; flag answers where it doesn't.

18 (Five compressions of one gradient). No single right winner; reward the ranking and defense. Typically the highest-bitrate 10-bit survives the most grade; the lowest-bitrate 8-bit bands first; HEVC/high-efficiency modes vary. The learning: a smooth gradient is the acid test, and it separates the formats fastest.

19 (Five profiles of one face). Model finding: the flattest/Log profile reaches the target grade most easily but looks worst ungraded; the standard profile looks best ungraded but resists a heavy grade. The point to land: "best ungraded" and "best after grading" are usually different clips — which is the whole standard-vs-Log choice, and why you pick based on whether a grade is coming.

20 (Read the pair). (a) 8-bit 4:2:0 = 256 tonal steps, quarter color — fine to watch/light-correct, bands under a heavy grade or key. (b) 10-bit 4:2:2 = 1,024 steps, half color — the sweet spot for serious grading and green screen. (c) 12-bit RAW = 4,096+ steps, full color, sensor data built in post — maximum latitude, maximum cost.

21 (Four briefs, four formats). Models: (a) same-day testimonial → standard, high bitrate, 8-bit 4:2:0 fine; no time/need to grade. (b) moody branded, graded next week → Log 10-bit (4:2:2 if available), high bitrate, exposed carefully; the grade needs the room. (c) green-screen product composite → 10-bit 4:2:2, high bitrate; the key needs clean color. (d) all-day conference, limited storage → efficient codec, a bitrate high enough to survive mixed light but sustainable for hours, fast offload; not RAW. Grade the reasoning, not the numbers; the answer to "why not RAW to be safe" is "the pipeline can't carry it and the job doesn't need it."

22 (Card-speed check). (a) 400 Mb/s ÷ 8 = 50 MB/s of actual writing. (b) V30 (30 MB/s) is not enough — it's below 50 and has no headroom; V60 (60 MB/s) clears it with some margin (V90 is safest). (c) If the card can't keep up, the camera stutters, drops frames, or stops recording mid-take — sometimes silently.

23 (Should I shoot Log?). (a) 8-bit phone, never graded → No. 8-bit Log bands when graded, and un-graded Log looks broken — both strings unmet; a standard profile will look better. (b) 10-bit camera, moody brief, a week to grade → Yes. All three strings met: they'll grade it, in 10-bit, with time to expose and handle it. (c) same-day wedding highlights → No. No time to grade; a standard profile delivers a finished look immediately.

24 (Write the offload plan). Model: fast card sized to the bitrate with headroom → offload to a laptop/SSD at the first break, not end of day → a second copy on another drive before either card is reused → open a clip and confirm it imports and plays → only then wipe a card. "Two copies before I reuse a card." Formalized as 3-2-1 backup in Chapter 37.

25 (Fix the won't-import file). Likely cause: a codec the software can't decode (a particular H.265 flavor or camera-specific codec) — not corruption, not the container/extension. Two fixes: transcode the file into a friendly codec (many free tools, or Resolve on ingest), or install the right decoder.

26 (Fix the banded sky). Two possible culprits: too low a bitrate and/or only 8-bit — a smooth sky is the hardest thing to keep and the easiest to band, and grading amplifies the damage. Should have shot the highest bitrate the card sustains and 10-bit if available. Damage is permanent; no post fix.

27 (Fix the grey, broken video). They shot Log (the "cinematic setting") and never graded it — and possibly in 8-bit. Two-part fix: immediate → grade it back to a normal look (color-space transform/LUT + a grade; Ch.31–32) if the footage holds up; next time → if you won't grade (or can't shoot 10-bit), shoot a standard profile, which looks finished out of camera.

28 (Fix the un-editable 4K). The clips are a heavy interframe/long-GOP codec (4K H.264/H.265): a media player streams them fine, but a timeline must rebuild every frame from neighbors, which chokes a modest computer. Standard fix: proxies / transcoding to an editing codec — lightweight stand-ins you cut with, swapped for originals on export — Chapter 27.

29 (Fix the lost footage). Two process failures: (1) the offload wasn't verified (they assumed it worked), and (2) the card was reused before a second copy existed. The rule that prevents it: never format a card until its footage lives in at least two other places and you've confirmed the copies open. This one can't be fixed in post — the answer is the habit. (3-2-1 backup: Ch.37.)

30 (Recreate the flat-to-graded reveal). Guidance: success = a clip that looks deliberately "wrong" (flat/grey) out of camera, built into a real, intentional look in the grade. Two sentences should explain that starting from flat/uncommitted footage gave more control than a pretty baked-in image, because the contrast and color hadn't been decided yet — the Fury Road strategy (Case Study 1) at your scale.

31 (Recreate the over-compressed look). Expect the student to name the artifacts they produced (macroblocking, banding, smearing) and one deliberate use — a "found footage," retro-web, surveillance, or degraded-memory aesthetic. Lesson: over-compression is normally a failure but occasionally a chosen texture; the difference is intent.

32 (Recreate the white-balance save). Expect: RAW (or the most flexible format) let them set white balance entirely in post with far more freedom than a baked-in standard clip, because the color wasn't committed at capture. The caveat they should name: this is a safety net, not an excuse to be sloppy — getting white balance close in camera still saves time and protects footage in less-flexible formats.

33 (Grade until it breaks). Guidance: success = documenting the exact point where the compressed clip falls apart (bands, blocks, smears) while the clean clip holds under the identical grade. That point, measured, is the ceiling of the capture format — the single most useful thing to know before a graded shoot.

34 (Transcode a stubborn file). Expect: transcoding to ProRes/DNxHR (or optimized/proxy media) makes it scrub smoothly — but the student must note it cannot add back quality the capture threw away, because transcoding re-packs existing data, it doesn't invent detail. Editing ease ≠ image quality.

35 (Build the grade the format allows). Expect: the high-quality clip reaches the intended look and pushes further before breaking; the 8-bit 4:2:0 clip hits its ceiling (banding/blocking) well short of the same look. The takeaway: when a heavy grade is coming, shoot the bit depth and bitrate that will survive it — decided on set, not in the edit.

36 (Format serves story). Model: "My Project 1 is an intimate testimonial I'll cut fast and correct lightly — so a standard profile at high bitrate, 8-bit 4:2:0, white balance set in camera; no heavy grade is coming, so latitude I won't use isn't worth the storage or the extra step." Weak answers can't connect the format to the post plan — flag those; arbitrary format is the habit this exercise cures.

37 (The irreversibility check). The recording format is most irreversible because you can re-light, re-frame, and re-cut, but you can never un-compress footage or add bit depth/range that was never recorded — the information is gone at the instant of capture. A pre-shoot test catches, for free, a won't-import codec, a card silently dropping frames, or 8-bit Log that bands the moment you grade — before the day that counts.

38 (Trace the pipeline end to end). A strong ~200-word trace names every stop and where information is kept or lost: light → sensor (grid of numbers) → demosaic (Bayer → full color) → picture profile (standard bakes in / Log stays flat / RAW skips it) → bit depth (tonal steps) → chroma (color detail) → bitrate/codec (compression) → container → import → grade. The permanent stops to mark: the profile (clipping/latitude), bit depth, chroma, and bitrate — everything the codec throws away is gone. RAW is unusual in deferring the profile/demosaic decisions to post.

39 (The three-projects format page). Model: talking-head → standard high-bitrate 8-bit (fast cut, light color); documentary short → Log/flat 10-bit if you'll grade the interviews and B-roll, else high-bitrate standard; branded piece → 10-bit Log (4:2:2 if keying), high bitrate, carefully exposed (heavy grade coming, and the pipeline to do it). Reward students who justify differences from each piece's post plan rather than copying one format everywhere.

40 (Teach "the video is not the footage," format edition). A strong ~200-word explanation shows that even though this chapter is all about the footage's technical form, the numbers only matter as room for the decisions post will make: 10-bit isn't "better," it's more room to grade before banding; Log isn't a look, it's information held for a later decision. The point to land: format is how you keep options open for the edit — so "the video is not the footage" still holds, because the footage's value is measured by what the edit can do with it. Flag any answer that calls a format "better" with no "in order to do X in post" attached.

Chapter 4 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises (numbering matches the final exercises.md, 1–40). These are models, not the only right answers — a defensible attempt that reasons from the chapter is worth more than matching this text. Shooting exercises are graded on whether the student can name what changed and why.

A. Seeing the lens

1. Spot the compression. Tells of a long lens: the background looks large and close behind the subject, out-of-focus even at a distance, space "stacked" and flat; shot from far away. Tells of a wide lens: the background is small and deep, edges may bow, the subject was close to camera, near-to-far reads sharp. The single best tell is how big the background is relative to the subject — big and close = long/far; small and deep = wide/close.

3. The "alone" lens. Look for a long lens from a distance: the character isolated, the background compressed into a wall, often shallow focus separating them further; the camera watches from far and never closes the gap. The loneliness is manufactured by the lens keeping us at a distance too — we observe rather than share. A wide lens would put us inside the space and destroy the isolation.

5. The rack in the wild (model). A strong answer names (a) the two focus planes (e.g., a cup foreground → a face in a doorway), (b) the motivation (a sound, a line, a movement justifying why attention travels now), and (c) what a cut would have lost — a cut tells us to look; the rack lets us discover the second subject the way we do in life, within one continuous, intimate shot. It only works because a shallow depth of field made "sharp" a place the eye is forced to follow.

B. Reading Described Shots

7. Name the fields (model). THE FRAME tells you both the focal length ("~85mm-equiv") and the aperture ("~f/2.2"). The focal length contributes the flattering compression of the face and the beginning of background separation; the aperture contributes the melted, anonymous background that erases the room and forces the eye to the subject's eyes. A good answer keeps these two contributions distinct — face rendering vs. room dissolving.

9. Diagnose the depth. The rack-focus reveal requires shallow depth of field because the whole effect depends on only one plane being sharp at a time — if both foreground and background figures were sharp, there'd be nothing to reveal and nowhere for the eye to travel. At f/11 the deep focus would keep both sharp; the figure at the window would already be visible, and the pull would do nothing. The director would have to reveal them another way — a cut, a camera move, or a light coming up on them. Shallow focus is what makes "in focus" a place a director can move.

B/C bridge

8. Rewrite the lens (model). THE FRAME (wide, close): "A person on a busy sidewalk, ~24mm-equiv from a step away; the street opens deep behind them, walls fanning to the sides, the far intersection tiny. We are on the sidewalk with them." THE EFFECT: "The open, deep space and the closeness put us inside the crowd beside the subject — we share the street rather than watch it. The long-lens isolation is gone; this feels immediate, involved." The relationship flips from observed and alone to present and among — same subject, opposite feeling, purely from lens + distance.

10. Write your own (model). Grade on: all seven fields present; THE FRAME correctly names the lens feel (wide/immersive vs. long/compressed vs. shallow/isolated) and ties it to subject-camera distance; THE EFFECT explains where the eye lands and why in lens terms; THE LESSON is a transferable principle, not a description. The re-create note should give a concrete own-gear plan ("shoot from across the car park on my phone's telephoto to compress the background").

C. Focal length in the hand

11. Same size, four lenses (model self-critique). Success: the subject is genuinely the same size in all four (if not, the student changed framing instead of isolating the variable). The write-up should note the background shrinking and deepening toward wide and growing and compressing toward long. The "most cinematic" pick is personal, but the reasoning must be about what the background did, not "it was blurrier." Common error: too little background depth in the room, so nothing changes — reshoot into a corridor.

12. The portrait ladder (trade-offs). Expect: ultra-wide up close = distorted (big nose, bent edges), unflattering; wide = still a bit unflattering; normal = honest, natural; short telephoto = flattering, features gently compressed; longest/far = very flattering but you're far from the subject (harder to direct, more room needed). Most pick the short telephoto as most flattering. Lesson: "flattering" is a distance effect (standing back), and the lens is what lets you fill the frame from that distance.

13. Perspective, proven. The background object will be dramatically larger in the long-and-far version and small in the wide-and-close version, while the person stays the same size in both. That's the whole proof: the person didn't change size but the background did, so the change came from where you stood. One-sentence takeaway: "Perspective — the size of the background relative to my subject — is set by my distance; the lens just lets me keep the subject framed from that distance."

14. The immersion shot. The wide-from-inside version should feel "you are here" — present, surrounded, subjective, deep space; the long-from-the-doorway version should feel "watching" — detached, observational, compressed. If the student can't feel the difference, check that they got genuinely close with the wide lens (immersion needs closeness) and far with the long lens (observation needs distance).

15. The normal-lens honesty test. The ~50mm piece should look plain and believable — no exaggeration, no flattery. Right choice when you want the viewer to simply trust what they see: a sincere testimonial, a documentary subject, a "just the facts" explainer. Too plain when the story wants heightened emotion (go wide for immersion or long for isolation/flatter). The lesson: neutrality is itself a choice, and sometimes exactly the right one.

16. Compression as story (model critique). Grade on: real background depth to compress (a long corridor/street), a genuinely long lens from far back, a still subject against a moving/compressed background, and a one-sentence feeling ("surrounded yet alone," "watched," "trapped"). Weak versions use a short lens or stand too close, so nothing stacks. The best versions marry the compression to a story reason for the isolation.

D. Aperture and depth of field

17. Read the scale. Order most light / shallowest → least light / deepest: f/1.8 → f/2.8 → f/4 → f/8 → f/16. The two "backwards" facts: (1) a smaller f-number is a bigger opening and more light; (2) that bigger opening gives less depth of field, not more. Light and focus move together and both run opposite to the number's size.

18. The depth-of-field ladder. Expect the background to go from a soft wash at the widest aperture to crisp when stopped down. Uses: wide open to isolate / a premium portrait / erase a messy room; middle for balance and focus safety; stopped down when the whole subject must be sharp or the context matters. Phone: portrait mode ON ≈ the wide-open look (check the edges); "subject moved far from the wall" shows real optical separation.

19. Isolate the ordinary (model critique). Success: an ugly, cluttered spot becomes a clean, beautiful frame because all three levers were used together (open aperture + longer lens + close to subject, subject far from background), erasing everything but the one object into soft bokeh. Grade on whether the distractions vanished — that's composition by subtraction. This is the exact trick that rescues shoots in bad rooms.

20. Bokeh hunt. Pleasing bokeh: smooth, round, evenly-lit orbs with soft edges; a creamy, calm falloff. Busy/nervous bokeh: edgy or doughnut-shaped highlights, harsh outlines, a jittery background that competes. What helps: a longer lens, a wider aperture, and real distance between subject and lights. Lesson: it's not just that the background is blurred, it's how — that "how" is bokeh.

21. Too shallow, on purpose. The wide-open failure should show the classic problem: eyes soft while the nose or ear is sharp, or the subject drifting out of focus when they move. The stopped-down fix (two stops down) restores a focus margin that forgives small movement. Keeping both side by side teaches the ⚠️ lesson permanently: wide open is beautiful and unforgiving; stop down for insurance.

22. Settings drill: choose the aperture (model). (a) Cluttered office to erase → open, ~f/2–f/2.8. (b) Whole product sharp → stop down, ~f/8. (c) Run-and-gun dim hall, unpredictable movement → ~f/4 if light allows (focus margin for movement; open only as far as you must for light). (d) Landscape, near flower + far mountain → ~f/8–f/11 (deep focus, both planes sharp). Each answer must name the reason, not just the number.

23. The story-driven aperture. Grade on whether the student chose by story, not prettiness. The right answer varies: a maker whose tools prove the story wants them legible (stop down); a coach delivering hard truth wants everything gone but their eyes (open up). The key insight — mirroring Case Study 2 — is that the more beautiful wide-open frame can be the wrong one when the background is evidence.

E. Focus for video

24. Peaking on. The student locates focus peaking (often under "focus assist" / "MF assist"), sees the coloured outline on the sharp plane, and watches it travel as they rack by hand. The deliverable is knowing where the setting lives on their device. No peaking on their phone camera? That's the argument for a manual camera app.

25. Pull a focus (model critique). A good pull is slow, smooth, and motivated — it leads the eye without lurching or hunting. Grade on: subjects at clearly different distances; an aperture open enough that only one is sharp at a time; a reason the pull happens when it does. Failures: too fast (a jerk), hunting (over/undershooting), or unmotivated (the eye has no reason to travel). Note how the wide aperture that makes the pull possible also makes it unforgiving.

26. Autofocus stress test. Expect hunting when light drops or contrast is low, and jumping to a crossing object or nearer face. The valuable output is a personal list of "when my AF fails" — exactly when to switch to manual on a shot that matters. Lesson: modern eye-detect AF is excellent and not to be trusted blindly on an unrepeatable take.

27. The motivated rack (model). Success: the focus travels on a clear cue (the near subject falls silent, a voice/movement calls attention to the far subject), so the pull feels like the viewer chose to look. The one-sentence defence should name the motivation ("the focus racks to the door on the sound of the knock"). An unmotivated rack — focus wandering for no reason — is the failure mode to catch.

F. Diagnose and fix

28. Fix the distorted face. Cause: an ultra-wide lens held close exaggerates near features (nose/forehead balloon) and bends the edges — wide lenses are unflattering at close range. Fix: switch to the main or telephoto lens and step back to fill the frame from farther away. Why they're linked: the distortion came from closeness (forced by the wide lens), so the cure is distance (enabled by a longer lens).

29. Fix the flat interview. Three changes: (1) go longer — a ~50–85mm-equiv, backing the camera up — to flatter the face and start separating subject from background; (2) open the aperture (~f/2.8) to throw the cluttered room out of focus; (3) move the subject away from the wall so the background falls further out of focus. Together these turn a flat "webcam" look into a separated, premium interview — all lens decisions, no new gear.

30. Fix the soft hero shot. Sharpening in post fails because the back of the product was never recorded sharp — the detail isn't in the file, and sharpening a blur only makes crunchy artefacts. On-set fix: deepen the depth of field — stop down toward f/8 (the biggest lever), and/or step back with a slightly longer lens, and/or angle the product so its depth sits within the focus zone. Confirm with a punch-in that the whole label is sharp before rolling.

G. Recreate It

31. Recreate the compression shot (model). Grade on stealing the technique, not copying footage: a genuinely long lens (or phone telephoto) from far back, a background with depth to compress, a still subject against a moving/compressed background. The viewer's reported feeling ("watched," "alone," "tense") is the deliverable — if they felt the isolation, it worked. Weak versions stand too close or use too short a lens.

32. Recreate the phone "portrait" honestly. The honest-separation version (subject far from wall + close with the telephoto) usually handles edges better — hair, glasses, held objects — because it's real optical falloff, not a software guess. Portrait mode can smear wispy hair or blur the wrong plane. Grade the comparison: can the student point to a specific place the computed blur guessed wrong?

33. The budget dolly zoom (model). Success is a recognisable effect, not a perfect one: the subject roughly holds size while the background visibly warps. Grade on (a) a background with depth to warp, (b) evidence the two moves were opposed (walk in + zoom out, or the reverse), (c) a motivated beat and a sound cue. The most common failure is mismatched rates (subject grows or shrinks) — catching which move is too fast is itself the learning.

H. Interleaved

34. Lens + shutter + frame rate. Success = both the motion foundation (24 fps, 1/50, the 180° rule from Ch. 2) and a deliberate lens look are set in one clip, stated with reasons. Proves Part I works as a system: motion foundation and lens storytelling are independent decisions made together.

35. Story is the boss (model). A strong chain: (1) the feeling ("I want the viewer to trust and warm to my subject"), (2) the focal length that says it ("~50–85mm-equiv so her face is honest-to-flattering and lifts off the room"), (3) the separation the story wants ("moderate — ~f/4 — enough to lift her off the background but keep her workspace legible"). Grade the reasoning chain feeling → lens → aperture, not the numbers.

36. Crop factor in practice. For a 2× crop sensor, a 25mm lens gives the 50mm-equivalent look. The out-loud explanation: the smaller sensor "crops in" on the image circle, capturing a narrower slice, so a 25mm on a 2× sensor frames like a 50mm on full-frame. This is why the book quotes full-frame-equivalents (Ch. 2 §2.5).

37. Full Part I baseline (model). A complete answer specifies, each with a one-line reason: resolution + frame rate (Ch. 2, "4K/24 for a filmic web piece with reframe headroom"), shutter (Ch. 2, "1/50, 180° rule"), capture format (Ch. 3, "standard profile for fast turnaround" or "Log for grading latitude"), and focal length + aperture + focus plan (Ch. 4, "85mm-equiv at f/4, manual focus with peaking on the eyes"). The whole of Part I assembled into one deliberate setup.

I. Synthesis and the Frame Log

38. Teach it back (model). A strong ~200-word explanation: perspective — how big the background looks relative to the subject, how deep or compressed space feels — is set by how far the camera stands from the subject, not by the lens. The lens only changes how much fits in the frame; it matters because it dictates how far you must stand for a given framing. Concrete example: film a friend the same size on a wide lens (step close) and a long lens (back up); the background balloons on the long lens and opens up on the wide, though the friend never changed size — proof the distance did it. Story consequence: a lonely character is shot long-and-far so we watch from across a gap we never close; a frantic scene is shot wide-and-close so we're thrown inside it. Grade on whether the student credits distance (not the lens) and ties it to a feeling.

39. The lens self-audit. A reflective grade — no single answer. Look for honesty: a student who marks most old shots "1–2 on intention" and can say why (autofocus and whatever lens was on the camera) has understood the chapter's core reframe — that a lens is a choice, not a default. The kept number is a baseline for measuring growth by Chapter 40.

40. The Frame-Log lens habit (model). Good entries are specific and causal: not "nice shot," but "the interview was a short telephoto wide open — the melted background made her the only thing to look at," or "they racked focus to her face on the word 'you,' so my eye moved on the exact beat." Grade on whether entries name a lens cause for a felt effect — that causal habit is the engine of the Frame Log and, eventually, of the student's taste.

Chapter 5 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises. Attempt each before reading. For shooting exercises, the "answer" is a critique rubric — what a strong result looks like.

A. Seeing exposure

1. Find the clip. A strong answer names specific clipped highlights and judges intent. Example: "A vlog shot toward a bright window — the window is pure white with no view visible (clipped, and it looked like a mistake, because the shooter clearly wanted us to see outside). A music video's backlit hair had a blown rim (clipped, but intentional — a stylized glow). A phone clip of a beach where the sky was solid white (clipped, mistake — a held highlight would've shown clouds)." The skill: telling deliberate clipping (a stylized flare) from accidental clipping (lost information the shot wanted).

2. Where's the skin? Model: for each, a one-word placement ("bright" / "mid" / "dark/moody") plus what it served. E.g., "A cooking show — bright, upper range: appetizing, clean, trustworthy. A thriller interrogation — dark, low: tense, hidden. A news interview — mid: neutral, factual." You're pre-training the judgment the waveform will make exact.

3. Spot the auto-exposure breathing. Strong answer describes one instance precisely ("as the vlogger walked from a shaded street into sun, the whole image dimmed for about a second, then brightened — the camera re-exposing") and prescribes the fix: lock exposure (manual, or AE lock on a phone) so brightness stays put; if the light genuinely changes across the shot, ride it deliberately rather than letting auto ramp it.

B. The locked triangle

5. Name the pinned lever. (a) The shutter is pinned. (b) The 180-degree shutter rule pins it — set to roughly 1/(2 × frame rate) for natural motion (1/50 s at 24 fps). (c) You don't move it for brightness because it's doing a motion job; changing it makes motion strobe (faster) or smear (slower).

6. Three ways to lose a stop. (a) Close the aperture one stop — side fee: deeper depth of field. (b) Lower ISO one stop — side fee: none, if you were above native (if you're at native, this isn't available). (c) Add one stop of ND — side fee: none. If the shot's whole point is a soft background, choose the ND: it darkens without touching the aperture, so your shallow depth of field survives. Closing the aperture would destroy the exact look you want.

7. The trade-back. (a) f/2.8 → f/5.6 is two stops (f/2.8 → f/4 → f/5.6), so you lost two stops of light (down to one-quarter). (b) Give it back without the shutter by removing ~two stops of ND (cost: none) if you had ND on, or raising ISO two stops (cost: more noise), or adding light (cost: time/gear). (c) Reach for pulling ND first — no penalty; ISO last, because it's the only one that costs image quality.

8. Diagnose the dead end. They're not stuck — they've forgotten the video shooter's essential tool. The fix with no side fee is an ND filter: it removes light while keeping f/2 and 1/50 s intact. A sunny day at f/2 often needs around six stops (an ND64) or more. The alternatives ruin the shot: a faster shutter strobes the motion (breaks the 180° rule); a smaller aperture kills the shallow-background look they wanted.

9. Write the recipe four ways. All four land at the same brightness for a bright outdoor interview wanting a shallow background; the differences are what they cost: - A — f/2.8, native ISO, ~6 stops ND. Best. Soft background (goal met), cleanest ISO, shutter pinned. ND does all the darkening with no image cost. - B — f/5.6, native ISO, ~4 stops ND. Less ND needed, but the background is now less soft — you sacrificed some of the look you wanted. - C — f/2.8, ISO below native (if available), ~5 stops ND. Slightly less ND, but if "below native" throws away dynamic range, you've traded highlight latitude for nothing useful. - D — f/2.8, native ISO, and a faster shutter instead of ND. Worst. Correct brightness but broken 180° rule — motion strobes. Ranking: A > B > C > D. A keeps every creative choice and pays no penalty; D violates the chapter's central rule.

C. ISO and native ISO

10. Find your base. Look in the manual (or spec page) for "base ISO" / "native ISO," and whether "dual native ISO" is listed with two values. On a phone, open a manual app and note the lowest ISO it offers. A strong answer records the number(s) and, if dual, both — e.g., "native 800; second native 4000." You now know where your image is cleanest, which changes step 2 of the workflow from "lowest number" to "native."

11. The noise ladder. Critique rubric: at native ISO the shadows are smooth; two stops up you begin to see fine speckle in the darkest areas; at the top you should see obvious grain and shifting colored noise in the shadows, softer detail, and "sandpaper" skin. A strong write-up notes that noise appears first and worst in the shadows and is baked in — you're documenting exactly what you'd inherit in the edit. Keep the clip as your personal ISO ceiling.

13. The last-resort test. Strong answer records the final ISO and shows the priority order was honored — aperture opened first, light added if any, ISO raised only as far as needed. The three-sentence defense should hit: darkness is missing information (unusable); noise is a texture (usable); an editor can attempt light noise reduction but cannot brighten a black frame into detail — so "a little noisy but visible" wins.

D. The ND filter

14. Read the ladder. ND8 removes 3 stops (it's also "0.9"). A 6-stop filter is an ND64 (1.8). If f/2.8 at 1/50 s and native ISO is six stops too bright, an ND64 (6-stop) brings it into range. All three labels — stops, ND number, density — describe the same glass because they're just three measuring systems for one amount of darkness.

15. Hold the shutter outdoors. Critique: in the auto version the still frames look fine but any motion strobes/stutters — the camera sped the shutter to survive the light, breaking the 180° rule. In the ND version, with the shutter pinned at 1/50 s, motion has natural blur and looks "shot." A strong paragraph names the hidden cause (the fast shutter) and the fix (ND lets the pinned shutter survive daylight). Keep this clip as your proof.

16. The variable-ND sweep. You should find a usable range where the image darkens cleanly and evenly, then, near the maximum, the ugly dark "X"/cross and/or a color cast appear. Answer: note the two angles (usable vs artifact-onset) so you always stay in the clean range — this is why a good variable ND (or a set of fixed NDs) matters. With fixed NDs, confirm each drops the waveform by exactly its rated stops.

17. Fix the shot. What went wrong: they left the ND on when moving from the bright exterior into the dim interior, so the camera was six stops down indoors — forcing ISO 12800 and all that noise to compensate. The one-second fix they skipped: take the ND off on the way inside. Correct re-exposure order once inside: native ISO, open the aperture, add light — and only then raise ISO if still short.

E. Reading the scopes

19. Read the waveform. Underexposed: the trace is crushed on the floor (0) — shadows featureless black, faces muddy, detail gone. Good: skin sits in the upper-middle (~60–70), nothing piled at 0 or 100 — detail everywhere it matters, room to grade. Overexposed: the trace is pinned to the ceiling (100) — highlights blown to featureless white. The failure you can never fix: clipped highlights — that information was never recorded, so no post can invent it.

20. Expose to a number. Rubric: the first face's skin sits around 60–70 with nothing pinned; the screenshot proves it. The point of the second, different-toned face: its "correct" placement legitimately falls at a different spot — a deeper skin tone sits lower, a paler one higher — and forcing both to the identical number would over- or under-expose one of them. Strong answers state this: you meter this skin, not a universal value.

21. Zebras two ways. They should land near the same exposure if your subject is lit normally: the 95–100% zebra warns you off clipping the brightest highlight, and the 70% zebra confirms the skin is sitting right — and a correctly-exposed face won't be clipping. If they disagree a lot (skin reads 70% but something also clips at 100%), you have high contrast in the frame (like a bright window behind the face) and must decide what to protect.

22. Learn your false-color key. Rubric: the student correctly identifies their device's three key colors — clipping, correct skin, noise floor — by pointing at known-bright, known-mid, and known-dark objects, and writes them down. Then, exposing a face by false color alone, the waveform confirms the skin landed in the upper-middle. The learning: the palette isn't in any book — it's specific to the device — but once the three colors are memorized, false color is the fastest way to nail a face.

23. Read the sequence. Two problems: (1) the face is underexposed (skin at ~25, too dark — recoverable but will get noisy when lifted); (2) the window is clipped (pinned at 100 — fatal, detail gone). The fatal one is the window. Two on-set fixes: (a) reposition so the window isn't a bright wall directly behind the subject; (b) add light to the face and expose to protect the window (bring it just under 100) so both the face reads and the window keeps detail. (This is exactly the Café Scene interior problem for Ch.11/13.)

24. Which tool when. (a) Run-and-gun handheld: zebras — a fast highlight alarm you can see without studying a graph while moving. (b) Precise skin on a Log talking-head: false color — it places this skin tone exactly, and it sees through Log's flatness. (c) Checking a bright sky while composing: the waveform (or a 95–100% zebra) — it shows precisely how close the sky's trace is to the 100 ceiling. The skill is matching the tool to the moment; none is "best" for everything.

F. Exposing Log and protecting highlights

25. The flat-image trap. Predictable disaster: clipped highlights. Log records flat and grey to hold range, so brightening "until it looks normal" pushes the highlights off the top without the flat preview ever looking alarming — and clipped highlights are gone forever. They should have judged with a scope (waveform/false color) against the camera's published Log targets, or previewed with a monitoring LUT while still recording flat.

26. Protect the highlight. Rubric: in the clipped take, the bright element is solid white and pulling exposure down in the editor reveals nothing — grey/white mush, because no detail was recorded. In the protected take (held just under 100), you can pull detail back — clouds in the sky, a view through the window. The lesson in the student's own words: highlights you protect on set are gradeable later; highlights you clip are simply gone.

27. Expose to your camera's Log targets. Rubric: the student finds their format's published middle-grey and white-card values, places a grey card and white card at those targets using false color or the waveform, then notes where a real face lands in the same light, and writes the full repeatable recipe. (No-Log version: three sentences correctly explaining that Log's flat, low-contrast image hides how close the highlights are to clipping, so it can't be judged by eye — and that the waveform and false color, or a monitoring LUT, are the tools to use instead.)

28. The ETTR argument. Four sentences, roughly: (1) Noise lives in the shadows, so exposing brighter lifts the whole image off the noise floor. (2) That makes the shadows cleaner once you normalize brightness back down in the grade. (3) The hard limit: never brighten so far that important highlights or skin approach clipping — clipped highlights are unrecoverable. (4) Use it when the scene has manageable highlights and clean-shadow priority (an interior with no blown windows); don't when there's a bright sky/window with little headroom, or no time to bring exposure carefully back down in post.

G. The workflow, locked

29. Run the six steps. Check your recreated checklist against FIGURE 5.4: (1) lock foundation — frame rate + 180° shutter (pinned); (2) native ISO; (3) aperture for the story's depth of field; (4) balance brightness on a scope — ND if bright, open/light/ISO-last if dark; (5) place skin, protect highlights; (6) lock and note it. Under a minute is the target; if you're slower, the usual culprit is fiddling with the (pinned) shutter or judging by eye instead of the scope.

30. Match two shots. Rubric: exposed to the same targets from your notes (not by eye), the two angles should sit at the same brightness and cut together cleanly. If they don't match, the usual drift is that the light actually changed between setups, or you re-judged by eye and landed differently. A strong answer diagnoses which and confirms that noting the recipe is what made re-creating the exposure possible.

31. Five Ways: one face, five exposures. Discussion of each: (a) correct — warm and inviting, the default for most talking-heads; (b) one under — slightly moody, still safe; (c) one over — brighter, more "commercial," risks the highlights; (d) protecting the window, face a bit dark — natural, realistic, the window reads; (e) silhouette — mysterious, anonymous, dramatic. The Godfather connection: (d) and (e) are motivated-darkness choices in the Gordon Willis lineage — letting a face fall dark on purpose to create mood and withhold information (Case Study 1). The lesson: five exposures, five different stories, from one setup.

32. Settings drill — the four setups. Check against the chapter's Settings Box. Strong answers (at 24 fps / 1/50 s): (a) window-lit talking head — native ISO, f/2–f/2.8, ND only if the window is very bright, watch the window clipping; (b) bright-sun walk-and-talk — native ISO, f/2.8, ~ND64/6 stops, ND off when you go inside; (c) product — native ISO, f/5.6–f/8 for detail, rarely any ND, watch specular highlights; (d) dim event room — native then raise ISO as needed, aperture wide open, no ND.

H. Interleaved

33. Shutter, aperture, ISO together. Shutter: 1/50 s (180° at 24 fps, Ch.2). Aperture: something wide like f/2.8 for a soft background — which also shrinks the depth of field, not just brightness (Ch.4). Too bright outdoors: add ND, not a faster shutter — because the shutter is pinned by the 180° rule and speeding it up would strobe the motion. In short: pin the shutter, pick the aperture for the look, and buy the darkness you need with ND.

34. From capture to exposure. Two disciplines Log now demands: (1) judge on a scope or under a monitoring LUT, never by the flat grey eye-look; (2) expose to the camera's published Log targets and protect the highlights (place middle grey/white/skin deliberately). The one thing to never do: brighten Log by eye until it "looks normal" — that clips the highlights. The extra dynamic range helps because it gives the bright sky more room below the 100 ceiling before it clips — more latitude to protect.

35. Depth of field meets ND. Work it out: f/11 is already a narrow aperture letting in little light — several stops less than f/2.8 — so in bright sun at 1/50 s and native ISO you're much closer to correct exposure, and you may need little or no ND (possibly none). The lesson: ND is most essential when you want a wide aperture (shallow look) in bright light; when the shot calls for deep focus at a narrow aperture, the small opening is already doing much of the light-cutting, so ND matters less. ND need scales with how wide you want to shoot.

36. The full Project 1 pass. A strong shooting card reads like: "Concept: a barista on why they love the morning rush (Ch.1). Format: 4K, 24 fps (Ch.2), standard profile, offloaded to two drives (Ch.3). Lens: 50mm-equiv at f/2.8 for a soft background (Ch.4). Exposure: native ISO, ND as needed by a window, judged on the waveform — skin ~65, window held under 100, locked and noted (Ch.5)." The value: Part I now fits on one index card you can shoot from — which means you've taken command of the camera.

Chapter 6

Chapter 6 — Answers to Selected Exercises

Model solutions and critiques for the odd-numbered exercises plus the even-numbered items the chapter promised a model for. Yours will differ — grade yourself on whether you can say the reason for each choice, which is the whole point of the chapter.

1. Find the thirds. Most professionally-shot faces put the eyes on or very near the upper-third line — you should find roughly 7–9 of 10 there, essentially none dead-center with the eyes in the middle. The exceptions you find will usually be deliberate symmetrical/centered shots (a face square to the lens for confrontation or formality) — note that those are choices, not accidents, which is exactly §6.2 + §6.4. Takeaway: "eyes on the upper third" is the professional default, and the deviations are motivated.

3. Which way is the room? In competent footage the placement matches the look: a subject looking camera-left sits on the right third with the open space (nose room) to the left. The beginner clip you find with it "backwards" will have the subject jammed against the edge they're looking toward, the empty frame stranded behind their head — it feels claustrophobic and wrong because the eye follows the gaze and finds a wall. The fix is always the same: place the subject on the third opposite the look.

5. Weigh the frame. The correct answer names the heaviest element and judges whether it's the intended subject. Watch for the classic thief: a blown-out window or lamp in the corner that is brighter than the subject's face — brightness is the heaviest quality there is, so the eye goes there first even though the person is the subject. If the heaviest thing isn't the subject, the frame is fighting itself; the fix is to tame the bright competitor (expose it down — Ch.5) or reframe it out.

7. Name the placement. FIGURE 6.3: the subject is on the left third; the eyes are on the upper-third line (modest headroom, no wasted ceiling); the nose room opens to the right, into the open café, where they'll look up when the message lands. If you got the direction of the nose room wrong, re-read §6.2's placement rule: the space opens in the direction of the look, so the subject sits on the opposite third.

9. Critique a frame. Four problems, four fixes: (1) Dead-center placement → put the subject on a third. (2) Too much headroom → frame the eyes on the upper-third line; let the head sit near the top edge. (3) A bright lamp brighter/sharper than the face stealing the eye → this is a visual-weight theft; reframe it out or knock it down in exposure so the face is the heaviest thing. (4) A merger — the door-frame line running out of the skull → step sideways or change height so the background line clears the head. Bonus: reserve clean negative space for a name super while you're at it.

11. Off-center in one minute. Model self-critique: the off-center take should read as "poised" and intentional; the centered take as "flat" and static. If you can't feel a difference, check that you actually moved the subject onto a third (not just nudged them) and that the eyes are on the upper third in the off-center version. The one-sentence reason you write ("the third placement gives the frame a direction the centered one lacks") is the deliverable — the shot is just evidence.

12. The headroom ladder. Trade-offs: (a) too much — face sinks to the bottom half, reads amateur, top of frame wasted. (b) modest — the safe, correct default for a medium shot; eyes near the upper third. (c) tight — more intimate, more pressure; good when you want to be close to the subject. (d) big close-up cropping the top of the head — most intimate/intense; only works if the eyes stay on the upper third and the crop looks deliberate (not an accidental beheading). Best answer usually: (b) for a neutral shot, (d) when you want intensity — and the reason is what matters.

13. Nose room both ways. The version with the subject jammed against the edge they're looking toward feels "wrong" because the eye instinctively follows a person's gaze and wants somewhere for it to land; with no room in front, the look "hits a wall" and the subject feels trapped and about to leave the frame. One-sentence answer: a gaze needs room to travel into, so the open space belongs in front of the face, not behind it.

14. Build a layer. Model answer: the flat version (person against a wall, everything on one plane) reads like a passport photo; the layered version — shooting past a foreground element (a plant, a mug, a doorway, a shoulder) at a wide aperture — reads as a deep space the subject lives inside. The mechanism (§6.3): three planes at three distances trigger the eye's depth-reading, and a wide aperture softens the front/back so the sharp midground subject pops. If it didn't work, your foreground was probably too far from the lens (get it close) or your aperture too small (open up / use portrait mode).

15. Lead the line. The version where the line points at the subject funnels the eye straight to them and feels composed; the version where the line leads away actively drags the eye off the subject toward nothing, and feels wrong even though the subject is present. Lesson: a leading line is only an asset when you use your feet to aim its convergence at the subject — an unaimed line is a liability. Best submissions place the subject exactly at the line's vanishing point.

16. Five compositions, one subject. Discussion: (a) centered — inert baseline. (b) on a third — instantly more alive, has a direction. (c) leading line into it — the eye is escorted to the subject; adds momentum. (d) big negative space — isolates and dignifies; sets mood (calm/premium/lonely). (e) foreground layer — adds depth; subject feels embedded in a place. There's no single "right" ranking — the point is that your intended feeling decides the winner, and you can name why each serves or fights that feeling. That naming is the skill.

17. Five ways to balance. Model answer: a heavy subject on the left third can be balanced by (i) a lamp (small but bright = heavy enough), (ii) a window, (iii) a second person (heavy — a face — so place them small/distant), (iv) a patch of texture, or (v) pure negative space. The most "settled" is usually the one where the answering weight is clearly smaller than the subject but bright/interesting enough to hold its side — proving weight ≠ size: a small bright thing balances a large dull subject like a light child far out on a see-saw balances a heavier one near the middle.

18. Turn on your grid. The setting is typically under the camera app's settings as "Grid," "Guides," "Composition," or "Framing." Model answer just confirms you found it and shot one clip using it. The real deliverable is the habit: leave it on permanently. If your camera also offers aspect-ratio or level guides, turn those on too.

19. Frame the same subject three ratios. Model changes: 16:9 — play it laterally, subject on a side third with nose room across the width, room for foreground layers. 9:16 — get closer and play it vertically: eyes on the upper third, stack elements top-to-bottom, keep the subject out of the unsafe bottom/right where captions and buttons sit. 2.39:1 — exploit the width: strand the subject on one side in a long sweep of negative space, or place two elements at opposite ends. One main change each, each with a reason.

20. Write the framing plan. Model answers: (a) corporate talking-head, YouTube, name super — 16:9; subject on a third facing slightly across frame; eyes on the upper third; nose room in the look direction; reserve a clean lower third for the name super; clean background, no mergers. (b) vertical cooking Reel — 9:16; get close; compose top-to-bottom (hands/food centered in the safe zone); keep faces and key action out of the bottom/right unsafe edges; big legible captions in reserved space. (c) wide establishing storefront — 16:9 (or wider); use the building's lines as leading lines; symmetrical/centered if the facade is symmetrical (a motivated exception); leave sky/foreground as negative space, possibly for a title.

21. Safe-zone stress test. After overlaying the platform's real furniture, the common casualty is on-screen text or a face parked in the bottom third or the right edge, covered by the caption bar, the like/share buttons, or the progress bar. Fix: move all critical elements — faces, key graphics, your own text — into the central/upper safe zone, away from the bottom and right. Lesson: on vertical, you compose around the app's interface, not just the frame edges.

23. Fix the merger. Three camera-only fixes (you can't move the furniture): (1) step sideways a foot or two so the pole/line no longer lines up with the head; (2) change your height (crouch or lift the camera) to drop the offending line behind the shoulder or above the head; (3) change the angle / open the aperture so the background — and the merging line — falls soft and recedes, reducing its pull. For the tangent, a small reframe so the two edges clearly overlap or clearly separate (never just touch) kills it.

25. Fix the moving frame. The error (§6.6) is no lead room for the moving subject: they start at the far edge with all the empty space behind them, so they immediately walk out of frame. Fix for a locked-off shot: reframe so the subject enters with open space ahead of them (place them on the trailing third with the frame open in the direction of travel) and let them cross into the space. Fix for a moving-camera shot (preview of Ch.8): pan or track with them so you maintain lead room the whole way — always keep open space ahead so they're forever walking into frame, never chasing them from behind.

27. Recreate a symmetrical frame. Model answer: your recreation should match the reference's centering and symmetry, not its content — a doorway, a corridor, a face down the lens. The two sentences on why it works should land on: symmetry reads as formality/order/control (or unease/comedy), and it works here because it's total and motivated — the subject or space is about order/confrontation/formality, so the formal frame means something. If your version feels like a mistake, the symmetry probably isn't complete (one side is cluttered/unbalanced) or isn't motivated by the subject. (See Case Study 1 — this is the Grand Budapest Hotel lesson.)

29. Compose for the title. If the text box covers the subject's face, the frame wasn't composed for the graphic — the fix is to re-shoot with the subject on a third and a clean lower third (or side) reserved for the super. This is the literal meaning of "you shoot for the edit": the composition decision (leave negative space) and the edit decision (put the title there) are the same decision made in two rooms. If it fit cleanly, you already think like an editor while shooting — good.

31. The whole Part-I-and-II frame. Model answer names four reasons, one per choice: focal length ("I used a longer lens to compress the background and make the subject sit closer to it" — or "a wide lens to include the room" — Ch.4); aperture ("I opened to f/2.0 to soften the foreground and background so the subject pops" — Ch.4 §4.3); exposure ("I set ISO/aperture/ND and checked the waveform so the window doesn't blow and steal the eye" — Ch.5); composition ("subject on the right third, eyes on the upper third, nose room left, and the counter as a leading line into them" — Ch.6). Four choices, four out-loud reasons — "motivate every choice" across two whole parts. That's the standard for the rest of the book.

Chapter 7 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises. Attempt each before reading. For shooting exercises, the "answer" is a critique rubric — what a strong result looks like — not a single correct clip.

A. Seeing the grammar

1. Count the sizes. A strong log entry names a number and a pattern: e.g., "This two-minute scene used four sizes — a wide to open, then it lived in two mediums/OTS for the exchange, dropping to a close-up twice, once on each person, for the two lines that mattered." The lesson to record: good dialogue scenes usually run on surprisingly few sizes (often 3–5), deployed with intention, not on constant novelty. If you counted a size that appeared only once, ask what beat earned it — that's almost always the scene's emotional peak.

3. Where's the line? Model: pick a scene and state which side the camera lives on and how you know — "Throughout, the detective stayed on frame-left looking right and the suspect on frame-right looking left; the camera never flipped them, so I could always tell who was where." If you find a scene where it does flip, note whether it flipped on a camera move (legitimate, Ch.9) or a hard cut (a possible error, or a deliberate one to signal a power shift). You are training the reflex to see the axis of action in finished work.

5. Read the height. Model answer: Motivated example — "a low angle on a boss standing over a seated employee: the height puts us in the employee's diminished position and makes the boss loom, which is exactly the power dynamic of the scene." Suspected unmotivated example — "a low hero angle on a character just walking down a hallway, with nothing in the story calling for grandeur; it read as the camera showing off." The skill being built: separating motivated height (a reason you can state) from decorative height (impressive-looking but meaningless). When unsure, the eye-level version is your control group — imagine the shot at eye-level and ask what was gained by moving.

B. Reading Described Shots and Sequences

7. Critique the coverage. Sizes in FIGURE 7.6: (1) WIDE master, (2) MS favor customer, (3) MS/OTS favor barista, (4) CU customer, (5) ECU insert (hands/cup), (6) WIDE master again. The shot-reverse-shot pair is shots 2 and 3 (customer favored looking right, barista favored looking left — matched, opposite eyelines, one side of the line). The shot that exists purely to let the editor hide a trim / compress time is shot 5, the ECU insert of the hands and cup — a cutaway you can drop over any join.

9. Write your own Described Sequence. A strong answer is a 5–8 row table with real observations and a walkthrough that explains rhythm, e.g.: "The scene opens on a 4-second wide (establish), then alternates ~2-second singles for the argument, accelerating to ~1-second holds as it peaks, then a longer 3-second close-up on the reaction to end. The shortening holds build tension; the final long hold is the punctuation that says 'this is the moment.'" Grade yourself on whether you noted (a) where it establishes, (b) the size on the key beat, and (c) how hold length changes with emotion — those three are the heart of reading coverage.

C. Shot sizes and angles

11. Five Ways: heights/angles. Expected feelings (yours may differ, and that's fine — the point is that you feel a difference): eye-level = honest/neutral; high angle = diminished/vulnerable; low angle = powerful/imposing; hard frontal = confrontational/formal; profile = distanced/observed. For a confident testimonial, choose eye-level (trust, equality). For a moment of doubt, a slight high angle can subtly make the subject feel smaller and more exposed. The deeper lesson: you just proved that height and angle are content — they changed the meaning without changing a word.

13. The three-size single subject. Critique rubric for a strong result: (1) all three sizes are cleanly distinguishable (a real wide, a real medium, a real close-up — not three near-identical framings); (2) the full line runs in each pass, so the same words exist at every size (test: can you cut from wide to close-up on any word? You should be able to); (3) exposure, white balance, and eyeline are consistent across the three, so they cut without a jump. If any two sizes are too similar, you have variety without usefulness — spread them out. This exercise is your Production Checkpoint; keep the footage.

D. The line and the eyeline

15. Break it on purpose. What goes wrong: in the correct version, A looks screen-right and B looks screen-left, so they read as facing each other. In the line-crossing version, the reverse now has B also looking screen-right (or on the wrong side of frame), so across the cut the two seem to be looking the same way — as if both staring at a third person off-screen — and the sense of a face-to-face conversation collapses. Viewers who can't name the rule still feel that "something is off." Keep both clips side by side; this contrast is the most convincing argument for the line you will ever have. (The controlled, deliberate ways to cross the line — a cutaway, a move that carries us across on camera — are Chapter 9.)

17. Over-the-shoulder pair. Critique rubric: (1) the near shoulder in each frame is soft (a foreground edge), not a sharp wall blocking the subject; (2) the near shoulder is on the correct side — matching that person's position from the wide (if B was frame-right in the wide, B's shoulder anchors frame-right in the OTS favoring A); (3) eyelines are opposite and aimed at the real off-screen position; (4) both OTS shots are from the same side of the line. If the pair cuts together and the two feel genuinely across from each other, you've nailed the single most useful dialogue setup there is.

E. Coverage and the shot list

19. Cut This: assemble your coverage. Guidance: build it in the master-and-coverage order — open on the wide to establish, cut into the medium for the body of the action, punch to the close-up on the key beat, and drop the insert over one join to smooth or shorten. A strong result reads as a continuous 10–15 second scene with clear geography and no spatial confusion. Common self-noticed problem: a jump because two adjacent shots are too similar in size/angle (the "not-quite-a-match" that reads as a stutter) — fix by changing size or angle enough between cuts (a preview of the 30-degree rule, Chapter 9). Note how ordering the same pieces differently changes the feel — that's the freedom coverage bought you.

21. Cover the Café counter. Rubric: you left the location only after capturing all of FIGURE 7.7's rows (master, MS each way, CU, ECU insert, re-establish); every setup is on the room side of the line; the customer faces/looks screen-right and the barista screen-left in every shot. When cut, it should match the shape of FIGURE 7.6 — establish, reverse-shot the order, close-up the beat, insert, re-establish. If a piece is missing, you'll feel exactly what you can't do (can't establish; can't hide a trim). That felt absence is the lesson.

23. Write the shot list. Model (person receiving a package at the door):

  # | Size  | Angle / setup        | Subject & action                | Audio        | Notes
  --+-------+----------------------+---------------------------------+--------------+-----------------
  1 | WIDE  | eye-level, street side | master: courier hands over box | sync + tone  | ROLL WHOLE beat
  2 | MS    | favor resident        | resident opens door, reacts    | sync         | eyeline to courier
  3 | MS/OTS| over resident's shldr | courier holds out the package  | sync         | opposite eyeline
  4 | CU    | resident              | the small smile / surprise     | sync         | the beat
  5 | ECU   | insert, slight high   | hands signing / taking the box | pen scratch  | cutaway to trim

Grade on: a master that covers the whole beat; a reverse-shot pair on one side of the line; a close-up saved for the reaction; an insert; and consistent screen direction noted in the "Notes" column.

25. Fix the shot: the flat scene. Three problems: (1) no rhythm or emphasis — every moment gets equal weight, so nothing stands out; (2) unreadable faces/emotion — a wide from across the room can't show what the conversation is actually about; (3) nothing to cut to — no way to shorten, hide a stumble, or punctuate. Fixes, all coverage: keep the wide as the master, then add mediums favoring each person, close-ups for the key beats, and one insert — all from one side of the line — so the scene can be assembled with rhythm and emphasis.

27. Fix the shot: the wrong size. The sizes are inverted relative to importance, violating the Hitchcock principle (size should track importance). The devastating news — the emotional peak — is buried in a wide where we can't read the face, while a throwaway line gets the emphasis of an ECU. Re-assign: play the climax in a close-up (or push to ECU on the eyes) so we live inside the reaction, and drop the trivial earlier line to a medium or wide. The rule of thumb: your tightest sizes are a scarce resource — spend them on the beats that matter most.

F. Planning and diagnosis (continued)

29. Recreate the "size = importance" move. What makes it land: the contrast between the wider setup and the close-up, and the timing of the cut. Shoot a medium of a subject doing something ordinary, then a close-up on the exact instant of the meaningful beat (the realization, the decision, the tear), and cut to the close-up on that beat — not before, not after. The jolt comes from the size change arriving precisely when the story asks the viewer to lean in. If it feels flat, check the timing (you probably cut too early or too late) or the sizes (they may be too similar to register as a change).

H. Interleaved

31. Coverage meets exposure (Ch. 5). Why mismatched exposure across coverage hurts: if your master is exposed at one brightness/white balance and your close-up at another, cutting between them produces a visible jump in brightness or color — the audience is pulled out of the scene, and you either live with the flicker or spend edit time matching shots (Chapter 31) that you never needed to mismatch. Strong practice: set and lock ISO, aperture, shutter, and white balance before you start covering a scene, and don't touch them between passes. Matching shots you never mismatched is free.

33. The full pre-to-edit loop. Model reflection: "Plan got right: I listed an insert of the object, which I'd have forgotten otherwise. Glad I shot: the reaction close-up — I ended the cut on it and it made the whole scene. Wish I'd shot: a second, wider re-establish at the end; I had nowhere clean to exit, so the cut feels abrupt." The value here is the loop itself — noticing, concretely, how a planning choice paid off in the edit and how a missed piece cost you. That noticing, repeated, is how a shooter learns to shoot for the edit.

I. Synthesis

34. Teach it back (model, ~200 words). "A scene isn't a thing you capture in one go — it's a set of pieces you gather and then assemble. Picture two people ordering coffee. You don't just film 'the order.' You shoot a master — a wide of the whole exchange, start to finish — which establishes who's where and is your safety net. Then you shoot it again, closer: a medium favoring the customer, the reverse favoring the barista, a close-up for the moment that matters, and an insert of the cup. Every pass runs the whole exchange, so the pieces overlap — the same words exist at every size, which lets you cut from any shot to any other at any moment. The one non-negotiable is the line: keep every camera on one side of the imaginary line between the two people, so the customer always looks one way and the barista the other, and the space stays readable. Do all that, and in the edit you can build the scene — and build it more than one way. That's the whole idea: you don't shoot a scene, you shoot its pieces, for the edit."

Grade on: (1) the master-and-coverage method named and explained; (2) the line named as the thing that keeps it cuttable; (3) overlap mentioned as what enables free cutting; (4) one concrete example carried through.

Chapter 8 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises. Attempt each before reading. For shooting exercises, the "answer" is a critique rubric — what a strong result looks like.

A. Seeing movement

1. Move-spotting. A strong tally shows that most shots don't move at all — in a lot of well-made drama and corporate work, well over half the shots are locked off, and the moves that exist are mostly slow and small. The learning: movement is the exception, not the rule, even in polished work. If your own tally of a video you're studying is "everything moves," that's usually a sign of over-moving, not high production value.

2. Motivated or arbitrary? Rubric: a strong answer names a concrete motivation for the good moves ("the camera pushed in as she realized the truth — following her attention inward") and is honest about the arbitrary ones ("a slow drift on a static interview for no reason I can find — arbitrary"). The skill is refusing to credit a move with motivation just because the production was expensive; even good work sometimes moves out of habit. Being able to spot the unmotivated move in others' work is how you stop making it in your own.

3. The still-frame audit. Rubric: the student confirms the camera is locked, then lists the internal motion carrying the shot — a subject gesturing, hands working, steam, background traffic, shifting light, a curtain. The point they should reach: a still camera and a still frame are different things; a great locked-off shot is usually full of motion, just not the camera's. This kills the beginner instinct that "still = boring."

4. Push or zoom? Answer: the reliable tell is parallax. If, as the subject gets bigger, the foreground and background shift relative to each other — objects slide past the frame edges, the background's relationship to the subject changes — the camera pushed in (moved through space). If everything scales up together and flat, with the background seeming to compress/press forward and nothing sliding past, it's a zoom. A strong answer names the exact tell used for each of the three shots.

5. Handheld vs. gimbal in the wild. Rubric: for handheld, answers should land on words like immediate, raw, tense, subjective, "you are here," documentary, unsteady-on-purpose. For gimbal/Steadicam: smooth, floating, elegant, controlled, dreamlike, gliding. The key insight is that this feeling comes from the movement itself, independent of content — the same action feels urgent handheld and serene on a gimbal. That's the emotional choice of §8.4 in a nutshell.

B. Reading the sequence

7. Locked vs. moved. Model: the locked version (FIGURE 8.1) witnesses the reaction from a fixed, neutral distance — it reads as objective, honest, calm, and it lets the performance carry the moment. The push-in version (FIGURE 8.3) presses into the moment, funneling attention onto the face exactly as the news lands, so the identical performance feels intensified and inescapable. A context to choose the locked version: a documentary or journalistic piece where you want the moment to feel un-editorialized and true — or any moment where the performance is so strong that a move would only distract from it. (Also valid: a comedic beat, where stillness lets the deadpan land.)

8. Diagnose the reveal. Model: rewritten as two shots — Shot A, a locked close-up of the note; cut; Shot B, a locked shot of the person in the doorway. What's lost: the continuity of discovery. In the single tilt, we travel from the note to the person, so we feel we found them ourselves — it's intimate and it happens on one continuous breath. Cutting hands us the two facts separately and externally; the connection ("this note, that person") is stated by the edit rather than experienced as a rising reveal. The move makes it discovery; the cut makes it exposition.

9. Write your own Described Shot. Rubric: all seven fields present, but the graded field is THE MOVE — a strong answer names the move precisely (not "the camera moves" but "a slow push-in from medium to close-up") and states its motivation ("as the character makes the decision, so our attention narrows with theirs"). Weak answers describe that the camera moved without saying why, which is precisely the habit this chapter is trying to break. THE EFFECT and THE LESSON should tie the move to a feeling and a transferable principle.

C. Shooting the moves

10. Lock it off. Critique rubric: (1) the camera is genuinely still — no drift, no sway; (2) the horizon is level; (3) the frame is alive with internal motion (the action, light, or background moving) so it isn't dead. If the shot feels boring, the fix is a stronger frame or a better action inside it, not a camera move — that's the whole point of the exercise.

11. The slow push-in. Answer: what separates a good push from a fidget is (a) speed — slow, slower than feels natural; (b) easing — it begins and ends gently, no jerk; (c) it ends on a composition you'd be happy to hold, on the emotional beat; (d) focus holds through the move; and (e) it's motivated — it lands as the sentence peaks. A fidgety push is fast, drifts to a stop on nothing in particular, and happens for no beat. Compare your push to your locked safety: if the locked one is actually stronger, that's a real and useful finding, not a failure.

12. The motivated pan. Critique rubric: (a) the following pan keeps correct lead room (space ahead of the walker), eases in/out, and starts/ends on a composed frame; the viewer shouldn't notice the pan because it matches a head-turn. (b) The reveal pan holds long enough on the face to raise the question ("what are they looking at?"), then moves at the speed of curiosity to the answer, settling on a composition. Common faults: overshooting and correcting; panning too fast (strobing); ending on nothing.

13. The tilt that reveals. Rubric: the opening frame raises a question (a detail — boots, a note, a base of a tall thing) and the tilt answers it (up to a face, or up to reveal scale). Strong tilts ease in and out and end on a held composition. The learning is that a reveal move is only satisfying if the start frame genuinely withholds something the end frame delivers — a tilt from nothing to nothing is just a camera moving.

14. Handheld, two energies. Critique rubric: both versions should be controlled — the difference is deliberate. The tightly-braced version (elbows in, wide lens) has subtle life and reads as "quietly present"; the energetic version reads as "urgent, unsettled." The key realization: handheld is a dial, not an on/off switch — you choose the energy to match the story, and neither extreme is "correct." Nauseating, uncontrolled shake is a failure of both; that's not "energy," it's just unsteady.

15. The tracking shot / walk-and-talk. Model critique: a strong result keeps the subject the same size and on the same third throughout, with lead room ahead; the camera matches their pace so the framing stays stable while the background flows past. Smoothness comes from a balanced gimbal or a braced wide-lens "ninja walk," not from the arms. Sound is on a lav (not the on-board mic). Common failures — pace drift (subject shrinks/grows), focus hunting, footstep audio — are exactly the ones Case Study 2 walks through and fixes.

16. The one-move short. Answer: placement is everything. The single move should land on the most important beat — the line that matters, the reaction, the reveal — so it reads as "the piece leaning in on the thing that counts." Placed anywhere else, it either wastes its impact or fights the beat that deserved it. The lesson the student should articulate: the one move hits hard because everything around it was still. Contrast is what gives a move its power.

D. One subject, many moves

17. Six ways, one moment. Model discussion: (1) Locked — objective, lets performance carry; the safe, often-best default. (2) Push-in — intensifies; best if the moment has a rising emotion. (3) Pull-out — releases/isolates; good for an ending or a reveal of context. (4) Pan/tilt reveal — good if there's information to disclose on a beat. (5) Handheld — immediacy/tension; good for a raw or urgent moment. (6) Gimbal/slider — elegance/glide; good for a smooth "accompaniment" feeling. The ranking depends entirely on what the moment is — a quiet realization wants the push-in or the hold; a chaotic argument wants handheld. The exercise's real lesson: the "best" move is defined by the story, not by which is fanciest.

18. Five feelings from one move. Answer: speed and size change the push-in's meaning dramatically. A very slow push over a wide-to-medium range reads as gentle, dawning, tender — good for warmth or quiet realization. A faster, tighter push into a close-up reads as aggressive, cornering, dread — good for tension or bad news landing. Longer lenses flatten and can feel more claustrophobic; wider lenses feel more like physically stepping toward. Tender moment → slow, gentle, modest size change. Dread → faster, into a tight close-up. Same move, opposite feelings, all from speed and framing.

E. Fixing broken moves

19. Fix the searching pan. What's wrong: the move has no motivation the viewer shares — the camera is hunting for the subject in real time (overshoot, correct, settle), so we're forced to watch it think, which reads as amateur. Fix: decide the exact start and end frames before rolling, rehearse the move once, then execute it clean. Never discover the frame on camera. If the subject's position is unpredictable, use a wider shot so you don't have to chase.

20. Fix the seasick follow. At least three fixes: (1) Lens — switch to a wide lens; the 85mm is amplifying every tremor, and a wide would massively reduce the apparent shake. (2) Technique/tool — put it on a gimbal, or brace hard (elbows in, strap tension) and walk heel-to-toe with bent knees. (3) Duration — don't hold a shaky handheld follow for 90 seconds; cut it up with cutaways so no single unstable shot runs long enough to nauseate. (4, bonus) Give the eye an anchor — keep a stable subject in frame. Root cause: a long lens + unbraced handheld + long duration is the perfect recipe for motion sickness.

21. Fix the empty glide. Diagnosis: the moves are unmotivated — smooth, but with no destination or reason, so they read as showing off the gimbal rather than serving the product. Smoothness is not a purpose. Fix in terms of motivation: give each move a job — a reveal (glide to disclose a feature), an arrival (settle on the hero product), a follow (track a hand using it) — and, crucially, let the camera stop and hold on the important frames. A brand video that never stops moving has no emphasis; the holds are what let a product land. Often the fix is fewer moves, each earned.

22. Fix the broken whip. Why it fails: a whip-pan transition can't be faked by adding a software blur between two locked-off shots — a real whip transition requires the actual motion-blur streaks of the camera whipping, in matching directions, on both shots, so the eye can't resolve the seam. Two static shots have no blur to hide the cut inside, so the software "whip" looks like a mushy artificial smear. What should have happened on set: shoot the end of shot A as a fast whip in one direction and the start of shot B as a fast whip in the same direction; then in the edit, cut at the blurriest frame of each. The transition has to be shot, not added.

F. Settings and scenarios

23. Settings for a smooth move. Shutter: 1/50 s (a 180° shutter at 24 fps) — this gives natural motion blur so the moving image looks smooth rather than strobed; keep it pinned and don't speed it up to fix brightness (use ND, Chapter 5). Lens/aperture to hold focus through the move: a wider lens and/or a smaller aperture (more depth of field) keeps the subject sharp even as the camera-to-subject distance changes during the push — so you're not fighting a focus pull while also operating the move.

24. The move-choice drill. Model answers: (a) Interview, hardest line → a slow push-in (or hold still) — "as they say the hard thing, I press in so the viewer feels cornered with them." (b) Hero product → locked off or a small slider push/reveal — "the product is the subject; I hold so the eye rests on it, or push slightly to present it, but I don't swirl around it." (c) Chef in a busy kitchen → a tracking shot / walk-and-talk on a gimbal or braced handheld — "I travel with them so we feel taken through their world." (d) Documentary final shot, someone who lost everything → a slow pull-out — "I pull back to leave them small in a large, empty world; the release is the ending." Each answer must have the one-sentence motivation; that's what's graded.

25. Rig the shot with what you own. Answer (phone + $30): (a) **Locked-off** — prop the phone on a stack of books, a windowsill, or a $10 mini tabletop tripod; brace it and it's dead still. (b) Smooth 2-foot push-in — a cheap phone slider, or set the phone on a rolling office chair / skateboard and push it slowly an inch at a time, or take one slow braced heel-to-toe step. (c) Smooth walking follow — an inexpensive phone gimbal is ideal near that budget; failing that, braced handheld on the wide lens with the "ninja walk" and the phone's built-in stabilization on. The point: every move in the chapter is achievable for almost nothing; the discipline (wide lens, slow, braced, motivated) matters more than the rig.

G. Stealing from the masters

27. Recreate a reveal. Model critique: the recreation succeeds if a viewer learns something at the end of the move that they didn't know at the start — and reports feeling a small "discovery." Strong versions withhold cleanly (the start frame genuinely hides the payoff), move at the speed of curiosity, and end on a held composition. If the viewer didn't feel a reveal, the usual causes are: the start frame gave the payoff away; the move was too fast to register; or the end frame wasn't held long enough to land. The learning is that a reveal is about timing and withholding, not about the mechanics of the move.

H. A first taste of the edit

29. Marry a whip. Answer: a successful marriage yields one continuous-looking whip that hides the cut inside the blur. If it doesn't marry, diagnose in this order: (1) Direction — are both whips going the same way? Opposite directions won't fuse. (2) Blur — is each whip actually fast enough to streak? A medium-fast pan that only stutters has no blur to hide the cut in. (3) Cut point — did you cut at the blurriest frame of each? Cutting on a readable frame exposes the seam. (4) Speed match — wildly different whip speeds read as a bump. Fix whichever is off and re-cut. If you never shot true whips, that's the lesson: it has to be captured on set (see Exercise 22).

30. The movement-and-stillness edit. Answer: the sequence should feel like it lands on the one move — several still shots build a calm baseline, then the single motivated move (on the key beat) delivers the emphasis. If your friend points to that move as "where it landed," the edit worked, and it proves the chapter's thesis: a move gets its power from the stillness around it. If they don't single it out, likely causes: too many other moves diluting the contrast, or the move placed on a beat that didn't deserve it. Tighten to one move, on the right beat, and try again.

I. Bringing it together (Interleaved)

31. Move + shot size. Answer: the same push-in means different things depending on where it starts. A push that starts on a wide travels a long emotional distance — from "here is the whole situation" toward "here is this one person" — so it reads as narrowing in, discovering the subject inside a context. A push that starts on a medium covers less ground and reads as a tighter, more intimate press onto someone we're already with. The medium-to-close-up push usually feels more intimate because it ends nearer the face and the whole move happens in the register of the personal; the wide-to-medium push feels more like revealing who the scene is about. Same move, different jobs, set by the size you start from.

32. Move + lens + exposure. Answer: the wide lens will look dramatically steadier than the long lens for the identical handheld walk — a wide magnifies your shake far less, so it reads as "gently alive," while the long lens amplifies every tremor into "earthquake" (Chapter 4, §4.1). On exposure: outdoors and bright, keep the shutter pinned at 1/50 s (a 180° angle at 24 fps) and cut the light with ND, not a faster shutter (Chapter 5, §5.3) — if you let the camera speed the shutter to survive the brightness, your walking motion strobes and stutters. The write-up should note both effects: the lens changed the apparent steadiness; the pinned-shutter-plus-ND kept the motion natural.

33. Move + the 180° line. Answer: this is the collision Chapter 9 is built on. A tracking or panning move that curves around one subject can carry the camera across the 180° line between the two people — and when it does, screen direction flips: the person who was facing frame-right is now facing frame-left, and if you cut out of that move to a normal reverse, the space breaks and the two people seem to stop facing each other. Watching your own clip, you'll likely see the flip happen at the moment the camera passes "through" the line. The lesson (previewing Chapter 9): a moving camera can legitimately cross the line within a continuous shot (the audience follows the move), but you must know it happened, because it changes what you can cut to afterward. Movement and continuity are entangled — which is exactly why Chapter 9 comes next.

34. The full Project-1 movement decision. Model card: "Concept: one piece of advice from someone I trust, said to camera, worth 60 seconds of a stranger's time (Ch.1). Framing: medium, subject on the right third, correct headroom and lead room (Ch.6–7). Lens/exposure: 50mm-equiv at f/2.8 for soft separation; native ISO, ND by the window, shutter pinned 1/50 s, skin on the waveform's upper-middle (Ch.4–5). Movement: shoot locked-off (the safety) AND a slow push-in over the key line — chosen move: [locked / push-in], because [one-sentence motivation] (Ch.8)." The value the student should name: everything learned so far now fits on one index card they can actually shoot from — which is what "taking command" looks like heading into the light and sound of Part III.

Chapter 9

Chapter 9 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises. These are guides, not the only right answers; the point is the reasoning.

1. Find the line. In a well-shot dialogue scene, person A stays on (say) the left of the frame and looks screen-right; person B stays on the right and looks screen-left — across every size, all scene long. That's the 180° rule holding: because the camera never crosses the line between them, screen direction is stable and you always know who is talking to whom without thinking about it. The certainty you feel is continuity working invisibly.

2. Track the cup. In most professional scenes the level holds because a script supervisor guarded it (or the shots were engineered to match). If you caught it jump, you found a real continuity error — note how, once seen, it's impossible to un-see, which is exactly why detail continuity is guarded so obsessively (§9.5).

3. Count the cameras. Evidence of two cameras: cuts within a single answer that land on different angles (not jumps), trims hidden without B-roll, a consistent A/B angle pair. Evidence of one camera: jump cuts, or every trim hidden behind B-roll/cutaways, or the answer never cut mid-sentence. Either is fine — the point is learning to read coverage backward from the finished edit.

5. Catch a jump cut and judge it. The vlog jump cut reads as style because it's deliberate and, above all, consistent — every trim hops the head the same way, so the viewer's brain reads a grammar, not a series of accidents. The awkward one reads as a mistake because it's isolated and unmotivated, rupturing an illusion of continuity the scene was otherwise trying to keep. Intent + consistency is the whole distinction (§9.4).

6. Name the match (FIGURES 9.3/9.4). Consistent elements: (1) same subject and same bag on the same shoulder; (2) same left-to-right screen direction; (3) same window key from camera-right (light continuity); (4) same pace, both takes running the full walk; (5) overlap (both cover the whole action); (6) same side of the line. Six is a solid catch; five earns the point.

7. Break it on paper. Rewritten THE MOVE: "Locked off from the far side of the line." Rewritten THE EFFECT: "The subject now crosses right-to-left; cutting the wide (L→R) to this medium (R→L) flips their direction — they appear to about-face mid-stride and walk back toward the door." The error is crossing the line (§9.2).

9. Write your own matched pair. A strong answer engineers the match explicitly. Example (pouring coffee): the wide THE CUT reads "runs the full pour, hand entering from frame-left"; the medium THE CUT reads "starts before the pour and overlaps it; cut on the tilt of the pot." Both THE FRAMEs keep the pot entering from the same side (screen direction); both THE LIGHTs name the same key. Weak answers describe two nice shots that don't share an action, a direction, or overlap — i.e., two shots that won't actually cut.

10. The reference-photo habit. Most people miss one small object or its exact orientation on the first try — which is precisely the point: memory is unreliable for detail, and a three-second photo beats it every time. This tiny gap between what you remember and what the photo shows is the entire argument for the reference-still habit (§9.5).

11. The seamless walk-in (self-critique). If the cut carried: you matched pace and screen direction and left overlap — congratulations, you cut on action. If the body jumped: most commonly the tempo differed between takes (fix: count the cross), or you cut a beat before/after the movement instead of on it (fix: find the frame mid-stride), or you didn't overlap enough (fix: run the full action in both sizes). Rank your miss against the five matching-action habits.

13. Obey the 30-degree rule. The zoom-only pass produces cuts where the subject "jumps" a few inches — same angle, too-similar framings. The move-around pass cuts cleanly because each size is a genuinely different viewpoint. The lesson lands physically: don't just zoom, move. Changing size is not enough; you must change angle by ~30°+ too (§9.4).

14. The crossed-line demonstration. (a) Reshooting from the correct side is the only full fix — the mismatch is gone. (b) A neutral head-on shot between the two softens it (the neutral shot belongs to neither side, so it bridges them). (c) A cutaway resets the viewer's spatial map so the far-side shot is partly forgiven. Most learners rank (a) best, (b)/(c) as partial rescues — which teaches that crossing the line is cheap to prevent on set and expensive-to-impossible to fix after.

15. The two-camera interview. With two angles ~30° apart, the mid-answer trim cuts to a genuinely different view and is invisible — the pro standard. With one camera and a reposition, the same works if you moved ≥30° and matched framing enough. The takeaway: a second angle is a license to trim; without it, every trim is a jump you must hide with B-roll or embrace as style.

16. Cut on the action. The on-the-movement cut is invisible; the before/after cuts stick out. Why: during the movement the eye tracks the motion and is briefly blind to the shot change, so the cut hides inside the motion; with no motion at the cut point, there's nothing to hide behind (§9.3, "Why It Works").

17. Hide the jump. (a) A cutaway/insert over the join hides the trim completely — best for material meant to feel continuous or polished. (b) Embracing the jump as rhythmic works when the piece wants fast, direct, talking-to-you energy (vlog/explainer). The right choice depends on the tone you're after; both are legitimate, and the wrong-for-the-material choice is the only real error.

18. Rescue a mismatch. You can usually hide a momentary detail mismatch by cutting away before the offending frame or choosing the matching moment or overlaying B-roll — but a mismatch that's present throughout (jacket open the whole close-up) can't be hidden, only avoided. Documenting what you couldn't fix is the lesson: detail continuity is far cheaper on set (a reference photo) than in post (a lost shot, or an unfixable error).

19. The invisible assembly. Success looks like a viewer not noticing the two cuts — establishing wide → cut-on-action to medium → insert. If they spot a cut, diagnose it against the checklist: mismatched pace, cut off the action, a screen-direction slip, or a light/exposure mismatch between takes.

20. Fix the shoot. The error is crossing the line (the hallway shot reversed screen direction). The two-word fix: same side — reshoot the medium from the same side of the line of motion as the wide. (No edit trick fully rescues it.)

21. Fix the jump. Ranked: (1) two cameras / two angles ≥30° apart — cut the trim to the other angle; invisible; costs a second camera or a reposition. (2) Cutaways / B-roll over the join — hides it; costs you having shot the B-roll. (3) Embrace the jump cut as visible/rhythmic — free, but only right if the piece's style welcomes it. The worst option is leaving a lone, unmotivated jump in a piece meant to feel continuous.

22. Fix the detail. On-set fix: match the cup's level between takes (a spare filled to the same line), or shoot all full-cup shots first, or decide the level and hold it — and take a reference photo. Edit-side rescues: cut away before the level is visible, or use the take/moment where it happens to match. On-set was far cheaper — the reference photo would have cost three seconds; the edit fix costs a shot or a compromise.

23. Write the continuity plan. Model: Line — camera on the window side of the tea-making path; screen direction — subject moves left-to-right from kettle to table; shooting order — wide of the whole action first (full overlap), then medium from ~40° around and closer, then insert of the pour; consumable — the tea level: shoot the full-cup shots before it's drunk, or use a stand-in cup; reference photo — the table's start state (mug position, kettle, spoon) before the first reset. A strong answer names all five; a weak one lists shots without the geography or the consumable plan.

25. Recreate a cut on action. Success = your cut lands mid-movement and the motion carries so a viewer doesn't catch the size change. If it bumped, you likely didn't overlap the action in both sizes or the pace differed. The content doesn't matter; the invisibility of the seam is the grade.

27. Five Ways: one entrance, five continuities (expected ranking). Typical most→least disorienting: (b) crossed line and (d) mismatched screen direction are the worst — spatial reversals read as the room flipping. (c) mismatched pace breaks the cut-on-action but is less globally disorienting. (e) mismatched detail (bag on the other shoulder) is a nagging "something's off" rather than full disorientation. (a) correct cuts clean. The exact order can vary, but spatial errors should top the list — geography errors hurt most, which is why §9.2 comes first.

29. Project 1 checkpoint (strong vs weak insert). Strong: the subject's hands doing something relevant (holding the tool they're describing, on the keyboard), lit from the same direction as the main shot, with the hands moving so it can cut on action, and several seconds with handles. Weak: a random detail shot lit differently (so it looks like a different room), static with no motion to cut on, or too short to lay over a trim. The test: will this lay cleanly over a cut in the main talking-head? If yes, it's the shot that makes the edit possible.

31. Frame + continuity (Ch.6, 9). Lead room (the space you leave in front of a moving/looking subject, Ch.6 §6.2) and screen direction reinforce each other: leaving lead room in the direction of travel both composes the shot well and visually states the screen direction, helping the match read. They don't fight — honoring lead room in both the wide and the medium makes the matched pair feel even more continuous, because the subject "leads" the same way in both.

33. Movement + continuity (Ch.8, 9). The moving camera is forgiven because it shows the crossing continuously — the viewer watches the geography rotate and updates their mental map in real time, so no reversal registers. A hard cut across the line gives no such transition: one frame the subject moves L→R, the next R→L, with nothing to explain the flip, so it reads as an about-face. The principle: the viewer accepts a change in screen direction if they're carried through it, and rejects it if they're cut into it (§9.2).

Chapter 10 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises (plus the even ones marked "answer/model/guidance provided"). These are models, not the only right answers — directing has a wide correct zone. Attempt each before reading.

A. Seeing the direction

1. Spot the eyeline. Model: A typical set — (a) a cooking YouTuber talking to the lens → direct address, feels like they're talking to you, a friend showing you something. (b) A TV news interview, subject looking just off to a reporter → off-lens, feels overheard, more objective/observational. (c) A movie character looking at another character → into the scene, we're a fly on the wall. (d) A vlogger walking and talking to a held phone → direct address, intimate/confessional. (e) A talking-head testimonial looking off to an unseen interviewer → off-lens, "real person telling their story." The point: the same shot size and lighting can feel completely different depending only on where the eyes go. Eyeline is a relationship choice.

3. Watch the tone. Model: In good behind-the-scenes footage you'll see directors doing tone work constantly — keeping their voice light and unhurried, laughing, saying "that was great, let's just try one more" (never "that was wrong"), using the performer's name, staying physically calm even under time pressure. The effect: the performers stay loose and keep offering ideas. Where a director is visibly tense or curt, watch the performers tighten up and stop volunteering. This is §10.1's claim — the set borrows its temperature from the director — visible in the wild.

5. Block the scene you're watching. Model (for a café scene where one person crosses to join another at a table): a top-down sketch showing person A seated frame-right at a table, person B entering from the door frame-left, the line running between the two once B sits, and camera positions — a wide establishing both, a medium favoring A, a reverse medium favoring B — all on one side of the line. The cut lands as B sits (a motion to cut on, §9.3). Grading yourself: did you notice the movement was staged to generate the cut? Did you keep all your inferred cameras on one side of the line? If a director had blocked B to enter from the other side, the whole coverage plan would flip. That realization — blocking dictates the line and the coverage — is the win.

B. Reading the direction

6. Name the beat. The four beats of FIGURE 10.3: (1) eyes drop to the phone; (2) the message lands (a flicker across the eyes); (3) one held breath; (4) eyes lift, out the window. It must be small because the push-in (Ch.8), the light, and the cut are already amplifying the moment — the performance's only job is to be true, and small, specific behavior reads as true while a big "shocked" face reads as manufactured. A director who lets the craft carry the amplification frees the human to simply react.

7. Diagnose the eyeline. Two problems with a camera on a low table (subject looking down into the lens): (1) Power/reading — a low angle grants the subject stature/distance; for a warm testimonial you wanted equal, and instead they read slightly superior or aloof. (2) Eyeline connection — the subject's eyes are angled down, so on screen they seem to be looking at something below the viewer rather than at the viewer/interviewer; the connection is subtly broken. One-move fix: raise the camera (and the interviewer) to the subject's eye level.

8. Rewrite the direction. Model rewrite of "really smile and show me how passionate you are!": "Forget the camera for a sec — tell me about the day you decided to open this place. What was that morning like?" (a true circumstance / memory question), and if you want an action: "Explain it to me like I'm a customer who's never been here and you want to win me over" (action verb: win me over). The smile and the passion now arrive on their own, because the person is doing something real instead of performing a state.

9. Write the reaction, undirected (model).

FIGURE (model) — "The reaction, badly directed"   [constructed teaching example]
  THE FRAME    Same medium at the window — but the face is doing a big, generic "shocked" expression: eyebrows
               up, mouth open, held too long. Eyes flick to the lens for approval.
  THE MOVE     Same push-in — but now it amplifies a fake, so it makes the fakeness *bigger*, not better.
  THE LIGHT    Same window key (light was never the problem).
  THE SOUND    A theatrical gasp that doesn't match a real person reading a text; over-performed.
  THE CUT      Hard to cut — the expression is so "on" there's no true moment to land on.
  THE EFFECT   We don't believe it. The bigness reads as acting; the glance to the lens breaks the world.
  THE LESSON   Directing the result ("look shocked") + a camera move that amplifies = a louder fake. The tell
               is in every field: the held face, the mismatched gasp, the approval-seeking glance.

The point: each field carries a tell — the held expression (frame), the amplified fake (move), the mismatched sound, the uncuttable moment, the broken world. Contrast with FIGURE 10.3, where every field is quiet and true.

C. Directing others

10. The two-minute relax (model self-critique). A strong write-up names the specific directing move that unlocked the best ten seconds. Model: "The most relaxed stretch came around 1:20, right after I asked her about the first bike she ever fixed and then said nothing when she paused. The memory question got her out of her head, and the silence let her add the line about her dad's garage — which she clearly hadn't planned and which is the best thing she said. What I'd change: I 'mm-hm'd' over two good lines; next time I react with my face only." If your best moment came right after you gave a task, asked a real question, or let a silence sit — that's your directing working, and naming it is the whole exercise.

11. Block a mark and hit it. With a mark, the subject lands in frame, in focus, in the light every take, and the takes cut together (matched positions). Without a mark ("stand wherever"), predictable failures: they drift out of focus, end up badly composed (too central, cramped, or with a pole growing out of their head), stand in a dark patch or blow out against a window, and — because their end position differs each take — the takes won't match for a cut. The lesson: a mark is not fussy; it's how you guarantee the technical craft you set actually contains the performance.

13. Direct the café reaction (guidance). Success looks like a small true reaction, not a big performed one. Guidance: decide privately what the message says and make it specific and real ("your sister just texted that she got the job you've both been anxious about for a week"). Tell the subject only the circumstance, not the feeling. Let them actually read something on the phone. Roll many takes; early ones will be performed. Watch for the take where a real flicker crosses their face and the breath is involuntary — that's the one. If every take is big/fake, you're probably still leaking a result-direction ("react when you read it") — strip it back to pure circumstance and silence.

15. Fix the direction. Three changes to "sit dad down, tell him to be himself, hit record, ask about the company": (1) Don't announce "be yourself" or hit a hard "record now" — chat first, roll early, slide in (kill the freeze trigger). (2) Ask specific memory questions, not "talk about the company" — "what made you start it?", "tell me about your hardest year" — to get real, human answers instead of a stiff brochure recitation. (3) Give his hands/attention something and get the eyeline + height right — sit at his eye level beside the lens so he talks to you, off-lens. Bonus: tell him the first take is a rehearsal; keep rolling after "done" for the steal.

14. The eyeline, three ways (self-check). You should feel: (a) to-lens = they're talking to you, most intense/intimate; (b) off-lens = you're overhearing, more relaxed and "documentary"; (c) into-scene = they're a character, you're invisible. If (a) felt uncomfortable to shoot, that's normal — direct address is the hardest eyeline for a non-actor, which is exactly why §10.6 exists.

D. Directing yourself

15/17 (solo). 17. Steal your own true take. The best take is almost never take 1 and usually lands around takes 3–6: early attempts flush out the stiffness and the "presenting" voice, and by the middle you've stopped trying and started talking. Reviewing the batch once at the end (instead of chimping after each) preserves the flow that gets you there. If your best take was your last one, you may have kept going long enough to relax — good; if it was an early one, you may have over-rehearsed into stiffness. The meta-lesson: treat yourself exactly like a nervous subject — warm up, roll long, don't judge mid-flow, pick the relaxed one.

15. Separate the roles. Model observation: when you lock focus/frame/exposure first (as operator) and only then step in to talk, the performance is noticeably calmer because your mind isn't split. When students try to adjust the phone while talking, they drift out of frame, lose focus, and lose their place — proof that operator-brain and performer-brain compete for the same attention. Fix codified: do all the technical work against a stand-in at your mark, lock everything, then perform and trust it.

E. Recreate It

19. Recreate the "overheard" interview look (self-check). Most subjects are more relaxed talking off-lens to you than to the bare lens, because a human face is something they know how to talk to and a glass eye isn't. If yours wasn't, check: were you at their eye height? Close enough to the lens (a hand's width)? Actually engaged with your face, or looking at your phone screen? The off-lens eyeline only relaxes people if there's a warm, present person on the other end of it. This is a direct rehearsal for Chapter 19.

18. Recreate a piece of "business" (model). The version with a task (talking while tamping espresso, wiring a bike, shaping clay) is almost always more natural, because: (1) attention goes to the task, not the face (kills self-consciousness); (2) the hands stop fidgeting or clamping; (3) the task reveals character and gives the edit something to cut to (a cutaway). The "just sit and talk" version leaves the hands and the attention with nowhere to go, so the nerves have room to live. This is Case Study 1's "glove" scaled to your subject — find their glove.

F. Five Ways

20. Five circumstances, one line (discussion). The real feeling usually does arrive on its own when you give only the circumstance, because a specific situation triggers genuine behavior. Which circumstance produced the truest read varies by person — often the one closest to something they've actually felt. The lesson: you never had to name an emotion; you gave a situation and the human supplied the truth. That's the entire mechanism of directing non-actors.

21. Five ways to relax one subject (self-check). There's no universal ranking — the point is that subjects differ. For one person the prop is everything; for another it's the memory question; for a third it's the steal after "wrap." A director learns to read which lever this person needs and pull it. If you found one tool did nothing and another transformed them, you've learned the most important meta-skill: relaxation is diagnostic, not a fixed recipe.

G. Fix the Shot & Settings Drill

22. Fix the direction. Three changes to "sit him down, tell him to be himself, ask about the company for a few minutes": as in Exercise 15 — (1) roll early, no hard "action"; (2) ask specific memory/story questions instead of "talk about the company"; (3) eye-level off-lens eyeline + a task for the hands + "first take's a rehearsal." The through-line: stop instructing a result and start engineering a relaxed conversation.

23. Fix the blocking. "Walk around while you talk" in a room with a bright window produces: drift out of focus (no mark, moving subject), passing through a dark patch (unplanned light), silhouette against the window (backlit). Blocking fixes: (1) set marks — a start mark and an end mark, both in good light and in the camera's focus zone; (2) position the camera so the whole path stays in frame and roughly in focus (or block a simpler move); (3) end the move facing the light, not against the window, so they're keyed, not silhouetted. Better yet, question whether the walk is motivated at all (§8.6) — often "just sit and talk, well-lit" beats an unmotivated wander.

24. Fix the solo take. Cold, stiff, eyes flicking away, forty escalating retakes. Diagnose against §10.6 and fix in order: (1) Separate the roles — they're probably operating and performing at once; lock focus/frame/exposure first, then perform. (2) Give the eyes a home — a photo or teleprompter at the lens, and "talk to one person," to fix the flicking eyeline and warm the delivery. (3) Stop chimping — the forty escalating retakes mean they're checking and judging after each; roll long, do a batch, review once, pick the relaxed one. Root cause: they're treating themselves worse than they'd treat a nervous subject.

25. Settings drill — direct address vs interview (model). (a) CEO welcome to new hires → to-lens (they're addressing the new people directly), eye level, relaxation move: teleprompter at the lens + several takes to warm up (execs are often stiff). (b) Shy artist describing their process → off-lens to you (far more relaxing than the bare lens) or into-scene while they work, eye level, relaxation move: give them the work to do with their hands. (c) Dramatic reaction shot → into the scene (a character unaware of us), eye level or a motivated angle, relaxation move: feed the circumstance, keep it small, use the camera/edit to amplify.

26. Write the blocking plan (model). Strong plan for a Project 1 talking-head: "Subject seated at the kitchen table (chair = mark), 3/4 turned to the window (their key). Camera at their eye level, framed on the right third, ~5 ft back at a ~50mm-equiv, f/2.8, focus locked on the eyes. Eyeline off-lens to me, seated at lens height camera-left — I want the warm 'overheard' testimonial feel, not a hard address. Background: the far wall ~8 ft behind, so it falls soft. Their hands get their coffee mug and, when relevant, the tool they're describing. First relaxation move: chat + roll early." Grading: does it stage people and camera together, set a mark, choose and justify an eyeline, plan the depth, and give the hands a task? That's the whole §10.2 + §10.4 checklist.

H. Interleaved

27. Block for the line (Ch.7 + 9). Model: two people at a table, the line runs between them. Choose the near side. Mark person A frame-left (looking screen-right at B) and person B frame-right (looking screen-left at A) so their eyelines meet across the cut. Shoot the two singles from the near side; direct each person's eyeline to a real point (each other, or your hand where the other's eyes are for an eyeline-only insert). Confirm: cut A's single to B's single — do they appear to look at each other? If both seem to look the same way, you crossed the line or mis-aimed an eyeline. This braids blocking (10), the line (9.2), and eyeline (7.4/10.4) into one move.

28. Direct a move to cut on (Ch.8 + 9) (guidance). Block a motivated cross (subject stands and crosses to the window — motivated by "let me show you"). Direct it as matching action: same path, same pace, in a wide and a medium, both running the full move with overlap (§9.3), both from one side of the line. Then cut on the movement (as they rise, or mid-stride). Success = the cut disappears inside the motion. This is the Chapter 8 move + the Chapter 9 matching-action + the Chapter 10 blocking, revealed as one combined skill: you block a motivated move so it both means something and gives the editor a seam to hide the cut in.

29. The composed, directed frame (Ch.6 + 10) (discussion). The hard part is holding both at once: nail thirds/headroom/lead room and a relaxed performance. Common trade: students who lock the composition freeze the person into a pose; students who chase the performance let the framing drift. The pro move is to block for both — set a mark that lands the composition, then relax the person within it, checking the frame between takes rather than policing it during. Whatever you traded, naming it is the lesson: directing is the art of holding several things true at the same time.

30. Teach it back (model). ~200-word model: "You can't direct a result because feeling can't be produced from the outside on command. Actors have technique to fake it convincingly; a real person doesn't — so when you tell your uncle to 'look excited,' he makes an impression of excited, and we can all tell. With actors you add (give them intentions and let their craft translate); with a real person you subtract — you remove the self-consciousness that's making them stiff. The way you do that is to direct the cause, not the result. Instead of 'be excited,' you give the true circumstance: 'this is the news you've been waiting three days for.' Now the real feeling shows up on its own, because he's responding to something instead of performing a state. It also works better small than big: in the café scene, the camera pushes in and the light and the cut do the amplifying, so the person only has to be true — a flicker across the eyes, a held breath. Big, manufactured reactions read as fake; small, specific, real ones read as true. Direct the cause, keep it small, and get out of the way." Grading: uses the add/subtract inversion, a concrete cause rewrite, and the small-reads-true point.

Chapter 11 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises. Even-numbered items that mirror an odd one follow the same reasoning; where an even item was flagged "(Answer provided)" a short note is included. These are models, not the only right answers — a defensible, well-reasoned attempt is the goal.

A. Seeing the light

1. Find the key. Look for: the bright side of the face (that's the key's direction); the catchlight position in the eyes (a dot high-left means the key is up and to the subject's right); the shadow of the nose (points away from the key, and downward if the key is properly above eye level). Most polished interviews: a soft key ~45° to one side and above, with modest fill. Most vlogs/webcams: a flat, on-axis key (ring light or window straight ahead) — flat, no shape. The single most common "amateur" tell is a face with no visible shadow at all.

3. Catchlight hunt. The catchlight is a map of the key. A dot at "10 o'clock" in the iris = key up and to the subject's right; "2 o'clock" = up and to their left; dead-center = an on-axis key (ring light) — flat. A doughnut/ring shape = a ring light. Two catchlights = two frontal sources (or a source plus a bounce). Absent catchlight = key too far to the side/behind or too low. The exercise trains you to read a whole lighting setup from the eyes alone.

5. The lighting map, from a still. Model reasoning: bright cheek camera-left + catchlight top-left ⇒ key ▢/☀ at camera-left ~45°, up. Shadow cheek carries detail (not black) ⇒ a fill ○ camera-right, ~2 stops under. Bright edge on the hair against a dark background ⇒ a rim ▣ behind, opposite the key. Warm pool in the background ⇒ a practical ▽. Place [CAM] facing the subject. Your map will be a guess; the skill is inferring instruments from evidence (bright sides, catchlights, edges, pools).

B. Reading Described Shots

7. Diagnose the field (fill raised to equal the key). THE LIGHT rewrite: "A large soft key with fill brought up to match it, so both sides of the face are equally bright — no shadow side, flat and open." THE EFFECT rewrite: "The subject looks approachable, unthreatening, and clean, but loses the modeling and mystery; the face is fully readable and a little ordinary." That flat, high-key look flatters corporate, beauty, product, and comedy — anywhere you want everyone bright, safe, and unmenacing. It would gut the confessional intimacy of the original, which depends on shadow.

9. Write your own Described Shot. Grade yourself on THE LIGHT: did you name the key's direction (which side, how high) and quality (hard/soft, from the shadow edges)? Did you estimate the key-to-fill contrast in stops (deep shadow = 3+; balanced = ~2; flat = 0)? Did you check for a rim (bright edge on hair/shoulder) and practicals (visible sources)? A strong answer reads a setup you could rebuild; a weak one just says "nice lighting."

C. Building the setup

11. The one-light key, five positions. Expected findings: (a) frontal — flat, shadowless, no shape (avoid); (b) 45°+up — natural modeling, both eyes lit, catchlight present — the warm interview choice; (c) 90° side — half-lit "split," dramatic and tense; (d) backlit — silhouette/rim, mysterious; (e) under-lit — shadows thrown up, menacing/unnatural. The warm interview uses (b) because the shadow is gentle and believable; the thriller uses (c) or (e) because the shadow is severe or unnatural. The lesson is entirely in the shadow the position creates.

13. Build three-point one light at a time. Model answer: (1) key only ⇒ shape (the face gets dimension, but the shadow may be too deep and the subject sticks to the background); (2) + bounce fill ⇒ contrast/mood (the shadow lifts to ~2 stops, the face warms); (3) + rim ⇒ separation/depth (a bright edge peels the subject off the wall). Naming each addition by its job is the whole point — you should be able to say what you'd lose by removing any one light.

15. Motivate it. The version with a visible practical (lamp/window/screen) as the apparent source reads as a real place — the viewer accepts the light because they can see its reason. The version with the same light but no visible source reads as staged/"TV-lit," even if technically identical, because nothing justifies it. Lesson: motivation is often about what's in the frame to explain the light, not just where the instrument sits.

17. Five key angles. 0° flat (avoid) → 30° gentle modeling → 45° the flattering default → 60° stronger shadow, more drama → 90° split/half-shadow (dramatic). The "tipping point" from flattering to dramatic is usually around 60°, where the shadow starts to dominate one side. Most portraits live between 30° and 45°.

19. Five sources. Typical soft→hard ranking: window (softest, big/close) → lamp bounced off a wall (soft) → screen (soft but weak) → bare lamp (harder) → phone LED (hardest, tiny source). The soft sources flatter most because a large apparent source makes gradual shadow edges. This directly proves the "apparent size of source" rule — the physically small phone LED is hardest; the physically large window is softest.

D. Five Ways

21. Fix the silhouette (listed under E; see E.21). — (Exercise 21 in Section E.)

(Section D even items 18: five heights — the two you'd never use for a friendly interview are under-the-chin (menacing up-shadows) and straight overhead (dark eye sockets); both put shadow where daylight never would.)

E. Diagnose and repair

20. Fix the raccoon eyes. Cause: an overhead ceiling fixture keys from straight above, so the brow shadows the eye sockets. Cheapest fix: turn the overhead off and key from a window or lamp ~45° off-axis and slightly above eye level, so the light reaches the eyes and the nose shadow falls small and downward.

21. Fix the silhouette. Sorted by stage: (pre/staging) turn the subject so the window is a side/front key, not behind them — or move them away from the window; (production/light) if they must stay put, add a light or a bright bounce on the camera side to key the face to match the window, or curtain/reduce the window; (exposure, Ch.5) expose for the face (open up / raise ISO / drop ND) and accept the window blowing out, or reduce the window's brightness so both hold. The root fix is staging: never let your brightest source sit behind your subject unless you want a silhouette.

23. Fix the flat corporate look. The problem is too much, too even light — zero contrast, zero shape. The fix is to remove light, not add it: kill one of the two sources (or pull it far back) so one becomes a clear key and the other drops to a ~2-stop fill. "More cinematic" almost always means more shadow, and shadow is created by subtracting fill, not adding light. Removing light is the fix because shape lives in the shadow you let return.

F. Plan the light

25. Read the box. Rows and why each matters: Key (source + 45°/up position — sets shape and exposure); subject-to-key distance (sets falloff/background darkness); catchlight (life in the eyes); fill (~2 stops under — sets mood); rim (separation); overhead OFF (so the key is the only direction); background (let it fall darker / add practicals — for separation and depth).

27. Contrast by the numbers. "Gentle but not flat" = about 2 stops key-to-fill. Achieve it with a bounce on the shadow side, slid in until the shadow lifts to that level (closer = flatter, back = deeper). Verify with the waveform or zebras (Ch.5): read the level on the lit cheek, then the shadow cheek; a ~2-stop difference confirms it — don't trust your eye alone, which adapts.

G. Recreate It

29. Recreate the window key. Hardest to match is usually the quality (softness) and the contrast: matching the direction is easy, but getting the shadow edge as gradual (source big/close enough) and the shadow depth to the same ~2 stops takes tuning the subject's distance to the window and the bounce. If your version looks harder, your source is smaller/farther than theirs — move the subject closer to the window.

H. Think in cuts

31. Match the coverage. With the lights untouched, the wide and close-up cut together seamlessly — the mood holds. When you deliberately change the fill between takes and re-cut, the subject's mood flickers softer/harder on every cut, and it feels subtly wrong. Lesson: lock the key-to-fill contrast and shoot all coverage of a subject at it — matched light on set is what makes the intercut invisible in post (Ch.7 → Ch.26/31).

I. Interleaved

33. Light + frame. A correct answer puts the subject on a third with proper headroom/lead room (Ch.6) and keys them from the lead-room side, so the lit cheek faces the open space they look into. When light and framing agree, the shot feels balanced; when the bright cheek faces the frame edge (away from the lead room), it feels off. The two crafts must be planned together.

35. Light + move + motivate. Model: the move's reason — "a slow push-in as the subject reaches the emotional core of their answer" (Ch.8); the light's reason — "keyed from the window, camera-left, because that's the real source in the room." Both sentences must be sayable out loud. Design the push so it ends where the key models the face most strongly (e.g., as the subject turns slightly into the key), so movement and light peak together.

J. Synthesis

37. Teach it back (model, ~200 words). "Point a light straight at a face and you erase every shadow — and with the shadows gone, so is all the shape: the cheekbones, the brow, the depth of the eye sockets flatten out, and the face looks like a passport photo. That's the trap: beginners think lighting means adding light until everything is bright. It's the opposite. A camera reads the pattern of bright and shadow across a surface, and it's the shadow that tells the eye 'this is round, this is deep, this has form.' So a cinematographer's real job isn't to add light — it's to decide where the shadow falls and how deep it goes. Try it right now: turn off your ceiling light, and turn someone toward a window so the daylight rakes across their face from one side. One cheek goes bright, the other falls into gentle shadow, and suddenly there's a face there — dimensional, alive, cinematic. You didn't add a thing; you let the shadow come back, and the shape came with it. That's the whole secret. Stop asking 'how do I make this brighter?' and start asking 'where do I want the shadow?' — and you've started actually lighting."

K. Make it a habit

39. The pre-flight lighting checklist (model). A good five-line checklist: (1) Overhead OFF — is the key the only direction? (2) Key 45° off-axis and up — catchlight high in both eyes? (3) Fill with a bounce to ~2 stops under (unless you want drama) — shadow modeled, not black or flat? (4) Separation — does the subject lift off the background? add a rim or darken the background if not. (5) Motivation — can I point to where every source "comes from"? The line most people skip under time pressure is usually #4 (separation) or #5 (motivation), because #1–3 fix the face and it's tempting to stop once the face looks good — but the background and the believability are what separate "fine" from "produced."

40. The lighting Frame Log (guidance). There's no single right answer — the value is in the accumulation. Strong entries name the key direction, the contrast (roughly, in stops), and any rim/practicals; weak entries just say "nice lighting." By day 7, students typically report they can no longer watch an interview without clocking the key. That automaticity is the whole goal; it's the same eye that will let you walk into any room in Chapter 13 and instantly find the softest light to put your subject in.

(Even-numbered items 2, 4, 6, 8, 10, 12, 14, 16, 18, 22, 24, 26, 28, 30, 32, 34, 36, 38 follow the same reasoning as their neighboring odd items; each was flagged inline in exercises.md with the expected finding.)

Chapter 12

Chapter 12 — Answers to Selected Exercises

Model solutions and critiques for the odd-numbered and ⭐⭐⭐ stretch exercises. Even-numbered items follow the same reasoning as their neighbors. These are models, not the only right answers — a defensible response that names quality, color, and ratio is the goal.

1. Hunt the shadow edge. You should find that flattering, "expensive"-looking faces almost always carry a soft shadow edge (a wide grey transition under the nose/jaw) from a big, close source, while gritty, dramatic, or cheap-harsh looks carry a sharp edge from a small/bare source. The reflex to install: judge light by the width of its shadow transition, not its brightness.

3. Rate the ratio. Expect comedy/commercial ≈ 1:1–2:1 (bright, open, safe), serious drama/thriller ≈ 4:1–8:1+ (shadow-dominated), corporate interview ≈ 2:1–3:1 (friendly but shaped). Yes — ratio tracks genre because contrast tracks emotional safety (§12.5). When a piece's ratio fights its genre (a flat, bright horror scene), it usually feels subtly wrong.

5. The white-balance hunt. A cozy café ad, a nostalgic flashback, or a kitchen-at-breakfast scene is almost always pushed warm (WB set below the source); a hospital, a morning, a thriller's "cold open," or a moonlit night is pushed cool (WB set above). The tell that it's deliberate: the warmth/coolness is consistent and flattering rather than a patchy, accidental cast. This is §12.3's core move — white balance as a mood dial, not just a correctness setting.

7. Diagnose the noir (FIGURE 12.7). Ratio: very high, roughly 8:1 or more — most of the frame is deep shadow. Quality: hard — you can tell because the venetian-blind slats throw sharp-edged stripes (a soft source would smear them into grey). Motivating prop: the (unseen) window with its venetian blind. What would destroy the mood: a strong fill light on the shadow side — it would lift the darkness, flatten the ratio, and turn menace into an ordinary evenly-lit shot. Low-key dies the instant you "fix" the shadow.

8. Write your own mood shot (model — "dread").

FIGURE (yours) — "The bad news"        [constructed teaching example]
  THE FRAME  Tight close-up, subject pushed to the frame edge, lots of dark negative space beside them.
  THE MOVE   Locked off; stillness holds tension.
  THE LIGHT  A single hard-ish key (~3200K) raked steeply from the side and slightly BELOW eyeline (unsettling),
             ratio ≈ 8:1, negative fill on the shadow side, background black. A faint cool rim for separation.
  THE SOUND  Bare room tone, one low drone; near silence.
  THE CUT    Holds a beat too long, past comfort.
  THE EFFECT Under-lighting + deep shadow + empty dark space = instinctive unease before a word is spoken.
  THE LESSON Dread is built from a high ratio, a hard/low source, and empty dark space the eye can't read.

Strong answers specify kelvin, ratio in stops, and a motivation for every source.

9. Bare vs diffused. The bare lamp gives a sharp shadow edge and shows every texture; diffused, the edge softens and the skin smooths. Use the diffused version for a friendly interview — soft light flatters and reads as approachable — reserving bare/hard for character, edge, or a "single hard source in the world" look. Same lamp, same place: the sheet is doing all the work.

11. Move the light, change the softness. Close, the softened source is larger relative to the face — softer, more wrapping, and it falls off faster (the background goes darker). Far, it's smaller-relative — a touch harder and more even front-to-back. Close usually flatters the face more and gives better control of the background (fast falloff separates subject from wall); far lights a scene more evenly. The lesson: distance changes both softness and falloff at once.

13. The gel look (model critique). A successful result reads as "a world," not "a filter": the color has a plausible off-screen source (a screen, a sign, a window) and it falls on the subject the way real light would — stronger where the pretend source is, wrapping and fading away from it. Common failures: the color is flat and even across the whole frame (no falloff → reads as a Photoshop layer, not light), or it's unmotivated (nothing in or implied by the frame could produce it). Fix by giving the color a direction and a reason. Bonus check: is the color clean or muddy? Muddy means a low-CRI source (§12.4).

14. Set a ratio by measurement. On the waveform, the key-lit cheek sits at some level; a 1-stop-darker fill side sits at roughly half that exposure (2:1), 2 stops ≈ a quarter (4:1), 3 stops ≈ an eighth (8:1). Confirm by eye: 2:1 looks "nice/natural," 4:1 "serious," 8:1 "moody." Most untrained taste lands around 2:1–3:1; deliberately pushing to 8:1 is how you discover you like more drama than you assumed. Braver-than-expected is the common finding — and a good sign.

15. Negative fill. The shadow side visibly deepens though you added no light — the black surface absorbs the ambient bounce that was filling it. Mood shifts from flat/safe toward shaped/serious. This is the key realization of §12.5: in a bright room, subtracting light creates contrast, and a piece of black board is a genuine lighting instrument. It feels backwards the first time and obvious ever after.

17. Five qualities, one face. For a beauty shot the ranking usually runs big window ≈ diffused ≈ white bounce (softest, most flattering) at the top, silver bounce (a touch harder/punchier) in the middle, bare lamp (hardest, least flattering) last. For a tough character shot the ranking often inverts — the bare hard lamp and silver bounce carve texture and edge that suit a rugged look, while the soft window reads as too gentle. The lesson: "best light" is relative to the story, never absolute.

18. Five ratios, one mood ladder. Side by side you'll typically feel "friendly" hold through 1:1 and 2:1, tip to "serious" around 4:1, and cross into "menacing/mysterious" by 8:1–16:1. Exact thresholds are personal and subject-dependent (a smiling face resists menace longer than a neutral one), which is the point: you now hold a dial for how safe the audience feels, calibrated to your own eye. Keep the five-frame strip as a reference.

19. The orange face. Error: white balance set to "daylight" (~5600K) under warm ~2900K lamps, so the camera under-compensates and leaves a strong orange cast. Fix: set WB to match the lamps (~2900–3200K — the bulb preset or the kelvin dial), or shoot a gray card and balance to it. If you like a little warmth, keep a touch — but make it a deliberate choice a notch off neutral, not an accident.

21. The flat, moodless "correct" shot. The single lever is ratio. The shot is technically perfect but lit to ~1:1–2:1 with full fill, so it has no shadow and therefore no mood. Fix: subtract the fill or add negative fill to push the ratio toward 4:1+, deepening the shadow side; optionally warm the key for intimacy. Mood lives in the shadow — give the shot some.

23. The muddy neon. Culprit: the low-CRI cheap RGB light. A poor spectrum has gaps, so when you demand a saturated blue there simply isn't clean blue to give — you get a greyish, desaturated approximation, and skin renders sick. Why: saturation ruthlessly exposes spectral gaps that white light hides. Fix: use a high-CRI (95+) source for creative color; if you must use the cheap light, keep the color gentler (less saturated) and check skin on a monitor. No grade fully rescues it — the color the source never emitted cannot be recovered.

25. Match the mismatch. Put a CTO (Color Temperature Orange) gel on the 5600K daylight LED to warm it down toward the 3200K lamp, then set white balance to tungsten (~3200K) so the whole scene agrees. (Equivalently, a CTB on the tungsten lamp cools it up toward daylight, then balance for ~5600K — either unifies the two; choose based on which look you want the room to have.)

27. Spec the shopping. Buy the CRI 96 bi-color panel, not the bright no-CRI one. CRI is the single spec you can't fix later — a low-CRI light renders skin badly no matter how bright it is or how you grade it — and bi-color plus dimming give you color-matching and ratio control that raw brightness never will. You're buying control, not watts (§12.6); a dimmer honest light beats a bright dishonest one every time.

29. Recreate high-key. You should end with a bright, shadowless, clean image: soft frontal key + strong fill to ~1:1–2:1, a bright even background, high-CRI light so skin reads healthy. Placed beside your #28 noir clip — same face, opposite genre — the only differences are ratio (low vs high), key hardness (soft vs hard), and background (lit vs black). That side-by-side is the chapter, proved with your own hands.

30. Warm-then-cool, one subject. Viewers should read the warm version as later-day/evening — cozy, safe — and the cool version as morning/night — clinical, tense. If they don't reliably distinguish them, the usual cause is that the cool version was flat (needs a slightly deeper ratio to feel intentional) or the white balance neutralized the color instead of preserving it. Success = color + ratio told the time and the mood with no words.

31. Soft vs hard in the grade. The soft-lit clip holds up: gentle gradients and intact highlights give the grade room to push contrast and warmth. The hard-lit clip breaks: blown speculars have no detail to recover and crushed shadows won't re-open. One-sentence lesson: you can add contrast to a soft image in post, but you can't remove the baked-in extremes of a hard one — so when unsure, shoot a little softer and shape it later (Theme 3, you shoot for the edit).

33. Two grades, one shot. You'll build a convincing warm/cozy and a cool/tense version from one neutral clip using temperature/tint + contrast. The argument for shooting in-camera instead: the low-key/high-ratio mood must be shot on set, because the grade can only push tones that exist — it can't invent deep shadow detail or true modeling the light never created. The warm/cool color shift, by contrast, is safely gradeable (especially from a flat/Log profile, 🔗 Ch. 3/31) — which is why color is the forgiving dial and ratio is the one to commit to on the day.

35. Light for the cut. The wide and the close-up should feel like the same moment — same key direction, same ratio, same color. If the light drifted (you moved the key, or AWB shifted the color between shots), the two won't intercut: the audience feels a "blink" of changed light at the cut even if they can't name it, and the editor is stuck. The fix is discipline: lock your WB, don't move the key between coverage sizes, and shoot the pair close together in time. This is shoot for the edit (Theme 3) applied to light — continuity of lighting is as real as continuity of action (🔗 Ch. 9).

36. Refine the Production Checkpoint (model self-critique). A strong result: the Chapter 11 version looks "lit" (correct but generic); the new version looks authored — you can name its mood in a word, its ratio in stops, and its white balance in kelvin, and every source is motivated. Common self-critique findings: (a) key wasn't soft enough — diffuse it more; (b) forgot the rim, so the subject muddied into the background; (c) left a touch of AWB drift — lock the kelvin; (d) played the ratio too safe — push it further than feels comfortable, because beginners reliably under-shoot mood. Three sentences should state the mood chosen, the ratio set, and what the color did.

37. Teach it back (model — ~200 words). "When you light a face, your instinct is to think about the light — where to put it, how bright to make it. But mood doesn't live in the light; it lives in the shadow it leaves behind. Picture two shots of the same person with the same key light. In the first, you bounce a white board on the dark side so both cheeks are nearly equal — a low ratio, maybe 2:1. It looks open, friendly, safe: high-key, the language of comedy and commercials. In the second, you change nothing about the key — you just put a black board on the dark side to soak up the spill, so that cheek falls to near-black. Now it's an 8:1 ratio: moody, tense, secretive — low-key, the language of noir and thrillers. Same person, same key light, opposite feeling, and the only thing that changed was how deep you let the dark get. That's why pros say you 'light the shadows, not the highlights,' and why a piece of black cardboard — negative fill — is one of the most expressive tools on set. Adding light is easy; deciding where the dark goes is the craft."

39. Mixed-source real room. (a) On one white balance, one source renders correct and the other throws a strong cast — a warm face against a blue window, or vice versa; usually unpleasant and hard to grade. (b) Unified reads clean and professional: gel the odd source (CTO/CTB) or reposition so one dominates, then balance to it — this is the safe default for most interviews. (c) Embraced can be the most beautiful if motivated: a warm subject against a cool window exterior is a real, filmable look (it's how much of Chapter 13's mixed-light material works). Best choice depends on story — unify for clarity/trust, embrace for mood — but never leave it accidental.

40. The product mood tie-in (model answer). Most people find the low-key "premium moody" version makes the object look more expensive: a single hard raking light reveals texture and form, the deep shadow implies exclusivity and restraint, and the high ratio reads as "considered." High-key clean reads as trustworthy and accessible (great for mass-market, food, or "friendly" products) but can feel generic/catalogue. The real lesson: there's no universal answer — luxury goods often go low-key and moody, everyday goods often go high-key and bright, and matching the lighting mood to the brand's price position is exactly the commercial instinct Chapter 22 builds on. Whichever you choose, a high-CRI source and a controlled ratio are what separate "premium" from "cheap."

Chapter 13 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises. Grade shooting exercises on reasoning and seeing, not polish.

1 (Where is the light?). The win is that students start naming direction automatically: "window, camera-left, from the side"; "overhead office light, straight down — dead eyes"; "lamp behind them, a rim on the hair." Push them past "it's bright" to a direction. This is the single reflex the chapter exists to build; twenty reps begins to install it.

3 (Read one room). A strong read names all three variables and a position: e.g., "One big window on the west wall — that's my key; I'll seat them beside it turned 45° toward it. It's soft because they're close to the glass. It's daylight, a bit cool, so I'll set ~5600K. The ceiling lights would fight it, so I'll switch them off." A weak read says "the light's fine here." Reward a decision, not a description.

5 (Name the color in kelvin). Grade on the instinct converging toward the table: lamp/candle ~2000–3000K, golden hour ~3000–4000K, midday ~5500K, overcast ~6500K, shade ~7500K, blue hour ~10,000K+. The learning is that "warm" means low kelvin and "cool" means high — counterintuitive at first — and that naming the number is the step before setting white balance for it.

7 (Diagnose the field — remove the bounce). Without the bounce, THE LIGHT loses its fill: the sun still rims the hair from behind, but the face — no longer having light thrown back into it — falls into deep shadow, potentially a near-silhouette. THE EFFECT shifts from "warm, glowing portrait" to "moody, dark, faceless figure against a bright sky." Neither is wrong; the bounce is the dial between "portrait" and "silhouette." The lesson: at golden hour the sun is the backlight and your bounce is the key — remove it and you remove the key.

9 (Write your own Described Shot). Grade on precision in THE LIGHT: does the student name the source (window/sun/sky/lamp), the direction (side/back/45°/top), the quality (hard/soft, read off a shadow edge), and a kelvin guess? A strong answer: "soft window key from camera-left, ~5600K, far cheek ~1.5 stops down, catchlight in the near eye." A weak one: "nice natural lighting." Specificity is the skill.

10 (The window portrait). Model self-critique: "The 45° turn gave the face real shape and a clean catchlight; exposure for the face is right; but the far cheek is a bit dark — next time I'd add a bounce — and I can see a hot blown strip where the sun caught the window edge, so I'd diffuse the glass." Names one concrete strength and one fixable flaw. The setup itself (window beside, 45°, expose for face) is the pass/fail core.

11 (Angle and distance). Expected findings: distance changed falloff/contrast — right at the glass the near cheek is much brighter than the far (high contrast, fast falloff); six feet back the two cheeks are closer in brightness (even, but dimmer). Angle changed which cheek is lit and how deep the shadow side is — facing the window lights both cheeks evenly; turned away deepens the shadow side. The lesson: two free "dials" (distance and angle) give full control of contrast without touching a light.

13 (Open shade rescue). Expected: the direct-sun version has squinting, dark eye sockets, and a hard nose shadow; the open-shade version (edge of the building, facing the open sky, shade WB set) is soft, even, and flattering, with light back in the eyes. Most students are surprised how much better a ten-foot move makes it — that surprise is the point: position is lighting.

15 (The no-budget three-tool day). Ranking usually runs: raw (worst — hard, top-ish, squinty) → +bounce (better, but still hard key) → +negative fill (shape returns) → +diffusion (best — the hard sun becomes soft). For most faces the diffusion makes the single biggest difference, because it fixes the root problem (a hard key) rather than patching its symptoms. That surprises people who expect the bounce to be the hero.

17 (Five directions from one window). No single correct pick, but the reasoning should note: facing the window = flat/even (good for beauty/youthful), 45° = classic portrait shape (usually the best all-rounder), profile = dramatic half-lit, turned away = deep shadow/moody, backed to it = rim/silhouette. The best pick for a friendly talking-head is typically 45° toward the glass — shape without losing the far eye. Reward the justification.

19 (Fix the silhouette). Problem: the subject is backlit by the window (shooting into the light), so the face silhouettes and auto-exposure drives it darker. Fastest fix: rotate the whole setup so the window is to the side (a soft key) and expose for the face; alternatively, if the window must stay behind, add light onto the face to compete with it. The one-line rule: don't shoot into the window; put it to the side.

21 (Fix the mismatched interview). Causes: (a) the sun moved and clouds passed over forty minutes, changing exposure and color; (b) white balance and/or exposure were left on auto, so the camera re-chose per take; (c) no attention to shooting coverage close together in time. Fixes by stage — pre: schedule a shorter window and pick a north window (consistent light); production: set manual, locked WB and exposure, shoot all sizes back-to-back, watch the window between takes; post: match shots in correction (Ch.31, §31.5) as a last resort. Meta-lesson: the cheapest fix is on set, by not creating the drift.

22 (Settings drill — golden hour). Model: WB manual daylight ~5600K (keeps the gold); 180° shutter; wide aperture (f/2.8) opening up as it dims; ISO low early, raised late; ND on early and pulled as the sun drops; expose for the face, protect highlights on the waveform/zebras; shoot the most important shot first. The ND explanation: it lets you hold a wide aperture while it's still bright, then comes off to recover exposure as the light falls. Shoot-order explanation: golden hour is a depleting, changing resource.

23 (Settings drill — the mixed room). Since the ceiling lights can't be turned off, the two live options are match (gel the window with CTO or the lamps with CTB, or re-bulb the fixtures to daylight, so everything is one color) or embrace (balance for the window key and let the warm ceiling lights read warm — only clean if they're not directly on the face causing a color split). A strong answer picks one, sets manual WB accordingly, and locks it; it explicitly rejects leaving the camera on auto WB. If the fixtures are fluorescent and greening the skin, matching/gelling or killing them is better than embracing.

24 (Settings drill — the café). Model, consistent with the locked geography: seat the subject by the window; on the reverse toward the counter the window falls camera-left as a soft daylight key across the face. Set WB manually for the window (~5600K), expose for the face, and embrace the mixed light — let the warm counter practicals glow as motivated amber pools behind. Do not turn the practicals off (they're part of the café's identity) and do not use auto WB. The result is the naturalistic warm/cool café look.

25 (Recreate the north-window portrait). Grade on whether the student matched the light logic: one soft directional key from a real window, a lit cheek and a gently shadowed cheek, a catchlight in the near eye, honest (not fully filled) shadow. They should report what a real window forced them to change (distance for falloff, a bounce for the shadow) — that problem-solving is the learning, not a pixel-match of the reference.

27 (Assemble the comparison). Guidance: intercutting the natural and built versions every few seconds exposes differences that side-by-side viewing hides — softness, color, noise, shape. A good write-up picks a winner for this piece and justifies it in story terms ("the window version's softness suits a warm, personal message; the built version's control would matter more if I needed to shoot at night or match across days"). The point is a defended choice, not "natural is better."

29 (Two moods from one shoot). Grade on whether a fresh viewer reads two different feelings from the same subject. Typically the midday/neutral clips + brisk pace read "ordinary/energetic"; the golden-hour + blue-hour clips + slow holds read "wistful/melancholy." The lesson lands when students see that they changed the feeling using light and pace alone, not content — the light is the mood.

31 (Light + exposure). Both exposures are valid for different stories: exposing for the face (background blows out) serves a portrait — we want to see the person, and a glowing white background reads as warmth/light; exposing for the background (face goes dark) serves a silhouette — mystery, anonymity, mood. The student should read each on the waveform/zebras and articulate that "correct" exposure is a story decision, not a single right value.

33 (The Production Checkpoint, defended). A strong defense compares on shape, color, cleanliness, and story, and commits: e.g., "I'll use the window version for Project 1 — its soft light suits the intimate, personal tone, it's cleaner than my one-lamp build, and the free daylight let me focus on the subject's comfort. I accept that it ties me to daytime shooting." A weak defense just says "the window one looks nicer." Require story-and-shoot reasoning.

35 (Frame Log, the light entry). Not graded for correctness; graded for the habit and for growing specificity. By week's end entries should move from "good lighting" to "soft window key camera-left, warm, far cheek in shadow — a north-window look." That drift toward precision is exactly the eye the chapter is building; it pays off in the case studies and in Chapters 29 and 39.

36 (The mixed-light decision). Expected findings: (a) killing the lamps gives the cleanest, most controllable face but a colder, emptier room; (b) embracing the mix and balancing for the window gives a warm, cozy, "real" look if the lamps are behind/around and not color-splitting the face; (c) balancing for the lamps makes the window go strongly blue — usually unusable unless you want a cold window as a design choice; (d) auto WB visibly hunts as the camera moves, and the takes will not match — the unusable option. Most students rank (b) most "real" and (d) worst. The lesson: decide and lock; never let the camera choose per shot.

37 (Recreate the Café look). Grade on execution of the geography and the WB decision: window as the soft daylight key across the face, a visible, motivated warm lamp behind as a practical, WB set for the window so the face is neutral and the lamp glows amber. It reads "cozy" rather than "wrong" because both colors are motivated — we expect a window to be cool and a lamp to be warm — so the warm/cool split matches the real world. If a student balanced for the lamp (window goes blue) or filled the whole face with the lamp (green/orange skin), that's the teachable failure: the mix is only beautiful when each color is justified and the key stays neutral on the face.

38 (The location audit). A strong audit reads each location in the chapter's terms and commits: e.g., "Location A: north window, soft/even, ~6000K, seat them beside it at 45° — consistent all day, my first choice. Location B: sun-facing window, hard/moving, would need diffusing — riskier. Location C: outdoor open shade at the building's edge, soft but ~7500K, needs shade WB and a bounce for the eyes — good backup." The winner should be justified on consistency and control for this piece, not just prettiness. The point is choosing the light in pre-production, so the shoot day is execution, not discovery.

Chapter 14 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises. Grade recording exercises on reasoning and hearing, not polish — and insist every judgment was made on headphones.

1 (Four-second audio test). The win is students noticing that sound often decides "stay or leave" before the picture does — a distant, hollow, or hissy voice repels within a second or two, while a close, present voice invites you in. Strong answers name a specific audio quality ("the voice was buried in room echo," "it was so close I could hear them breathe"), not "the audio was bad." This installs the chapter's core reflex: judge sound consciously, the way an audience judges it unconsciously.

3 (Room-tone safari). The learning is that no room is silent. Strong entries name the actual components of each room's tone — "the kitchen is fridge compressor plus a clock tick," "the bedroom is almost nothing but a faint street," "the bathroom is a low HVAC roar and its own hard-surface hollowness." The point that pays off in Ch.15: you must record this tone deliberately, and quieter rooms make cleaner dialogue — so choosing (or quieting) the room is a real production decision.

5 (Rate the audio, not the picture). Model critique: a phone vlog might score presence 4 (close, arm's length), noise 2 (roomy/AC), levels 3 (a bit jumpy); a broadcast piece scores 5/5/5. The intended surprise is that "expensive-looking" and "good-sounding" do not always travel together — plenty of high-gloss footage has mediocre, roomy sound, and plenty of cheap footage sounds great because someone got a lav close. That decoupling is the proof that audio is its own craft, won by placement, not by the camera budget.

7 (Proximity proof). Expected: the arm's-length take is thinner, more distant, more room; the six-inch take is present and warm — same mic. Strong write-ups attribute it correctly to the direct-to-reverberant ratio, not to "the close one was louder" (loudness is incidental; you can match levels and it's still cleaner). This ten-second clip is the whole chapter; students should keep it and replay it whenever they're tempted to buy a mic instead of moving one.

8 (Close, on-axis, out of frame). Model critique of a good submission: ranks built-in-far worst (roomy, distant), lav or close-held directional best (present), and names why — "closeness added presence and killed the room; aiming the directional mic at my mouth added crispness; when I let it point at the wall it went dull." The learning is that all three placements share one goal (get close and aimed) reached different ways, and that a healthy meter reading did nothing for the far take.

9 (Lav-rustle hunt). For each induced noise, the fix should be named precisely: collar rub → mount the mic where fabric can't touch it (and tape down the rubbing edge); cable swing/thump → tape a small strain-relief loop just below the capsule so cable movement never reaches it; off-center/head-turn dropout → center it high on the sternum (omni forgives the turn). The meta-lesson students must state: none of these is fixable in post — a rustle sits inside the word — so the lav is a five-minute job done right or an unusable take done carelessly.

11 (Room-tone recording). Guidance: after laying the recorded tone under a gap between sentences, the "dead hole" of digital silence disappears and the gap sounds like the same continuous space. A strong reflection connects it forward: this is why Ch.15 makes recording 30 seconds of room tone an unbreakable habit — it's the connective tissue that lets an editor hide the seams between takes. If a student's tone doesn't blend, usually the tone was recorded in a different spot or after conditions changed (AC cycled) — same-room, same-time is the rule.

13 (Mic a second person). Model critique: with one mic serving two mouths, students discover the core difficulty of coverage for dialogue — a single lav favors one speaker; a mic you swing between them risks missing the start of a line or adding movement noise; a close-held mic must chase whoever's talking. The honest conclusion is that two mouths really want two mics (or a boom operator actively following the dialogue) — which is exactly the problem Ch.15's booming and Ch.19's two-mic interview solve. Reward students who felt the problem, even if their solution was imperfect.

14 (Re-read FIGURE 14.1). The two SOUND fields: Take A (built-in, far) = thin, distant, voice behind a wash of room echo and fridge/street, ends of sentences lost; Take B (close mic) = present, warm, voice in front of the room, breath and consonant edges audible, noise pushed to the background. The single change that produced the entire difference: the mic's distance from the mouth (nothing visual changed). That's sound is half the picture made audible.

15 (Write the SOUND field). Grade on precision in THE SOUND: does the student name the dialogue's presence (close/roomy), the room tone/ambience underneath, any music or effects, whether it's sync or off-screen, and audio's job in the shot? A strong answer: "close, dry sync dialogue on a lav; a low room tone of an office under it; no music; the closeness makes the moment feel intimate and true." A weak one: "good audio." Specificity about presence is the skill.

16 (Diagnose the mic from the sound). Strong reasoning works backward from what's heard: very present, consistent, slightly chest-heavy, immune to head-turns → a lav; open and natural with a touch of room and a slight overhead character → a boom; thin, roomy, distant, level drifting with framing → on-camera. The point is that presence + room + consistency are audible fingerprints of placement — you can hear where the mic was without seeing it.

17 (Rewrite a failure as a fix). Model: change the placement, then rewrite the fields. "THE ADJUST: swap the camera's built-in mic for a lav clipped on the sternum (or a boom just off the top of frame, aimed at the mouth). THE SOUND (new): close and present — the voice steps in front of the room; the fridge recedes to a distant background; every word of the trailing clause is clear. THE EFFECT (new): reads professional; the viewer settles in." The lesson: the fix is placement, made on set — not a post rescue.

18 (One line, five treatments). Discussion: the expected ranking usually puts (c) close lav and (d) well-aimed directional at the top, (b) built-in-close in the middle, (a) built-in-far and (e) mis-aimed directional at the bottom. The instructive surprise is (e): the expensive, capable directional mic, pointed at the wall instead of the mouth, can lose to a cheap built-in mic held close — because aim and distance beat capability. Students should state the general law: for a mic, being in the right place beats being the right price.

19 (Five rooms, one mic). Expected ranking: the clothes-filled closet and carpeted bedroom win (soft surfaces absorb reflections → low reverberation → clean); the tiled bathroom and bare kitchen lose (hard surfaces → echo → roomy); outdoors is clean of reflections but may add wind/traffic. The lesson students must articulate: the room is part of your signal. A mic can't un-hear reflections, so choosing (or softening) the space is a recording decision — which is why voiceover artists record in closets.

21 (Fix the recording). Sorted by cause: placement — the voice is hollow/distant because the mic was too far; fix by getting a lav or boom close to the mouth. Room — the AC is as loud as the words; fix by switching it off during takes (and grabbing room tone) and by moving closer so the voice dominates it. Gain — one sentence crackles because a peak clipped; fix by setting gain to the loudest real delivery with headroom (peaks ~-12, never 0), possibly with a safety track. The meta-point: a gorgeous picture guarantees nothing about the sound — they're captured separately.

23 (Fix the levels). The -35 dBFS file is clean but far too quiet: the editor can raise it, but boosting also raises the hidden hiss (noise floor), so it gets noisier as it gets louder — recoverable, at a cost. The 0-dBFS-pinned file is clipped: the peaks are chopped flat and the waveform data is gone, so no software rebuilds it — generally unsalvageable. The lesson: too quiet is a recoverable mistake; too loud is fatal. When unsure, err low and protect headroom.

25 (Read the meter). A correct sketch marks: 0 dBFS at the top = the clipping ceiling ("never hit this"); ~-12 dBFS = where dialogue peaks should sit ("loud, clean"); ~-40 to -50 dBFS = where room tone/quiet lives ("fine down here"); and the span between your peaks and 0 = headroom ("safety margin for surprises"). The one-line takeaway: aim for -12, never 0, and keep headroom.

27 (Auto vs manual). In the pauses of speech, AGC hears "no signal" and cranks the gain up hunting for something to level — which lifts the room hiss into an audible swell; when speech returns it ducks the gain again, producing a queasy breathing/pumping under everything, and levels that don't match shot to shot. The two-word instruction: go manual. Set the level yourself against the subject's real, loudest delivery and leave it.

29 (Recreate a podcast's intimacy). Discussion: matching a podcast's intense presence usually takes getting very close (a few inches, off-axis to avoid plosive pops), a soft/dead space (soft furnishings, a closet, blankets — kill reflections), and a clean level. Students typically report they had to get much closer than they expected and quiet the room more than they expected. The lesson: "present and intimate" is manufactured by distance and a dead room, not by an expensive mic — the same levers as every other exercise, pushed to an extreme.

31 (Match or fail). Expected: cutting a close take against a far take of the "same" line produces an audible jump in presence and room level at the cut — the piece sounds broken even though the words are continuous. The lesson in the student's words should be: consistent mic placement on set is what makes an edit possible; if distance/aim wander, the takes won't match and no post fix cleanly reconciles them. This is you shoot for the edit, applied to sound.

33 (Sound for coverage). To keep sound matching across shot sizes (a wide and a matching CU, Ch.7): keep the mic in the same place even as the camera moves — a lav rides the subject so it never changes with framing (ideal); a boom must be re-set to the same distance/aim for each size; and levels/WB-equivalent (gain) stay locked. If you must change mics between sizes, the presence won't match and the cut will betray it. The rule: the camera can move for the frame, but the mic should stay put for the sound.

35 (Frame Log — listening entry). Not graded for correctness; graded for the habit and growing specificity. By week's end entries should drift from "good audio" to "hidden lav, very present, dry room, no music — the closeness sells the sincerity." That drift toward naming mic, presence, and audio's job is exactly the ear the chapter builds, and it feeds the case studies, Ch.29's edit sensibility, and the reel in Ch.39.

12 (Set a level from scratch). A correct result: gain set so normal speech peaks ~-12 dBFS, confirmed by delivering the loudest likely line and seeing it stay under 0; then deliberately over-driving it to hear the clip (a crackly buzz on the peaks). Students should record the working gain number and note that they set it against the loud delivery, not the rehearsal murmur — because people get louder on "record."

20 (Fix the plan). The baked-in problem: the camera across the office puts its mic far from the mouth, so the recording will be hollow and roomy no matter how good the picture. One-sentence fix: get a mic close — a lav on the subject or a boom just off-frame — instead of relying on the built-in mic at framing distance. (Bonus: switch off the AC and grab room tone.)

22 (Fix the shotgun). Two reasons it fails: (1) on the camera hotshoe, the shotgun is as far from the subject as the camera — too far — so the room floods in; (2) indoors in a tiled (hard, reflective) kitchen, the shotgun's interference tube gathers reflections and sounds hollow/roomy. Fix: get the mic close and off the camera — a lav on the subject, or a supercardioid boomed just out of frame (indoors a supercardioid beats a shotgun). Aim it at the mouth.

24 (Fix the lav). Two distinct problems: (a) movement noise — the mic is rubbing clothing and/or the cable is transmitting thumps; fix by mounting it clear of rubbing fabric and taping a strain-relief cable loop. (b) off-axis dropout — it's off-center, so turning the head dulls the voice and drops the level; fix by centering it high on the sternum. Both are on-set placement fixes; neither is repairable in post.

26 (Write the audio settings). Models: (a) seated talking-head → lav (wired), centered on the sternum, gain to -12, AGC off, headphones; (b) walk-and-talk outdoors, windy → wireless lav with a furry windscreen (wind protection is Ch.15), close on the chest, gain to -12; (c) busy market person-on-the-street → handheld stick held right at the mouth (visible is fine), or a very close lav, gain conservative for sudden loud noise; (d) two-person scripted from six feet → shotgun/supercardioid on a boom, close and just out of frame, aimed at whoever speaks (or two lavs). Reward mic-chosen-by-situation reasoning.

28 (Budget the two-mic safety). Model: clip a lav on the subject (consistent, always close — the safe track) and simultaneously boom a shotgun/supercardioid just off-frame (open, natural — the preferred sound), recording both. If the boom drifts off-axis, a battery dies, or a cable rustles, the other covers it; in the edit you choose the better sound per line and patch a missed word from the safety. Pros do this because a re-shoot is expensive and a lost line is unrecoverable — redundancy is cheaper than regret. (Connects to Harry Caul's multiple mics in CS-1.)

34 (Exposure/audio parallel). Model: judging exposure on a waveform/zebras and judging level on a dBFS meter are the same discipline — trust an instrument, not just your eye/ear, and respect a hard ceiling you must never hit. For picture the ceiling is clipped/blown highlights (the top of the waveform); for sound it's 0 dBFS clipping (the top of the meter). Both ceilings destroy information permanently, both are protected by leaving headroom, and in both you expose/level for the important thing (the face / the dialogue) while watching the peaks.

Chapter 15 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises in exercises.md (plus every even item that poses a checkable question). Self-directed shoot/perception items (2, 4, 12) have brief guidance only. Numbers match the exercise file.

A. Training the ear

1. Hear the room. You should find that the "quietest" room is often the loudest once you truly listen — a tiled bathroom rings and amplifies, a kitchen hums with a fridge, a "silent" bedroom hides HVAC and traffic. The lesson: your brain filters constantly; the mic does not. The carpeted, soft-furnished room is usually the best for dialogue because soft materials absorb rather than reflect.

2. The clap test. (Self-directed.) Guidance: the longest ring = the most reflective, worst room for dialogue (hard, parallel, bare). The most "dead" room (clothes, carpet, soft furniture) is easiest to record clean. This clap is the single fastest room-scout you can do — make it a habit on arrival.

3. Become the microphone. Typical finds: an HVAC or fridge drone, a computer or projector fan, traffic through the walls, a fluorescent/electrical buzz, and the room's own echo — all things your ears had tuned out. The point: what the mic captures is the truth of the recording, and sealed headphones make you hear it. This is the core habit of §15.3.

5. Diagnose by ear alone. Model diagnoses: (a) hollow, distant, "in a stairwell" → reverb; fix on set by getting closer + deadening the room. (b) steady low drone under everything → HVAC/hum; kill the motor, aim rejection at it, high-pass on. (c) thin and far, room louder than voice → mic too far; get the mic close (inverse-square). (d) scratchy on movement → lav clothing rustle; re-mount. (e) crackle on loud words → clipping; lower gain, keep headroom. Naming the fault is the skill; each maps to a specific on-set fix.

B. Reading Described Shots

6. Name the fields. The plan captures the dialogue with two lavs (one on each speaker) and a boom (favoring the talker); the fourth thing recorded — after the dialogue — is room tone. Two lavs and a boom because the lavs give a clean, close, isolated track on each voice through the noise, while the boom is the natural-sounding master; together they give the editor both isolation and a believable room sound, plus redundancy.

7. Critique the boom. Two faults: (1) held dead-center between two people — both voices sit off-axis and distant, so both sound weaker and roomier than they should. Fix: favor the talker, swinging the mic to whoever speaks and leading the next line. (2) level with their chests — chesty, boomy tone, and it risks a mic-in-frame or picking up table/hand noise. Fix: boom overhead, aimed down at the mouths, a hand's width out of the top of frame.

8. Read the meter. Dialogue averaging -6 dBFS with peaks at 0 means almost no headroom and the peaks are clipping. On the loudest word you'll hear a harsh crackle/distortion that cannot be repaired. Fix: lower the gain so the average sits near -12 dBFS and the loudest line peaks under -6, restoring 6+ dB of headroom.

9. Write your own (model).

FIGURE (model) — "Interview by a busy window, room deadened"   [constructed teaching example]
  THE FRAME    Medium of a subject on the left third, window camera-left; a blanket-draped wall behind.
  THE MOVE     Locked off on a tripod.
  THE LIGHT    Soft window key camera-left; far cheek a stop down (Ch.13).
  THE SOUND    Shotgun boomed overhead, a hand out of frame, aimed at the mouth: the voice close and dry
               because the mic is near and the blankets have killed the room's ring. Traffic hum outside
               sits low under the high-pass. Room tone (the faint street + HVAC) will bed under it in post.
  THE CUT      Cuts to a matching close-up; the room-tone bed hides the seam.
  THE EFFECT   The subject sounds present and intimate despite the busy room — proximity + absorption won.
  THE LESSON   You beat a live room by getting close and deadening it, not by fixing it later.

C. Recording on location

10. The scratch-and-clap (guidance). Success check: in your editor, you should see a tall narrow spike at the clap in both the camera's audio and the second phone's audio. If you can't see it in the camera track, the camera mic was off — that's the failure this exercise exists to catch. Slide one track until the spikes stack; lips and clean voice now match.

11. Boom it overhead (model self-critique). Grade yourself on three things: (1) evenness — does the level/tone stay steady, or swim? Swimming = you moved the capsule off the mouth; aim more precisely. (2) frame — scrub the top edge; if the mic dips in anywhere, you were too low — hold a hand's width above the frame line. (3) shadow — check the wall for a moving bar; if present, reposition off the key's throw. A good take: close, dry, even, no mic, no shadow.

13. Set levels the right way. The correct version: normal speech near -12 dBFS, the loud line peaking under -6, clean throughout. The too-hot version: the loud line hits 0 and crackles — permanently. A/B them and the lesson is visceral: the safe direction is down. Keep the ruined clip; it's the most persuasive argument for headroom you'll ever own.

14. Deaden a bad room (model). Expected result: the bare-room take sounds hollow and distant ("bathroom"); the deadened-and-closer take sounds close, dry, and present — often dramatically so, with the same mic. In words: the blankets absorbed the reflections that were smearing the voice, and moving closer raised the direct voice over what reflections remained. This one exercise disproves "I need a better mic" for most beginners.

15. Tame the wind. Expected: the bare-mic take has a low roar and thumps that swamp the voice; the protected take is markedly cleaner. Ranking of what helped: outdoors, the furry cover usually helps most, then body-blocking the wind, then a foam screen (foam alone is often not enough for real wind). Indoors/breath, foam is plenty. The high-pass helps with residual rumble but can't fix direct wind hits.

16. The full clean-dialogue take (model checklist-critique). Score the take against §15.3: clipping? hum? HVAC? wind? rustle? plosives? echo? off-mic? intermittent? dropouts? Then confirm: levels averaged ~-12 with headroom; room tamed if it rang; motors off; and you grabbed matching room tone. A pass = clean on all ten faults and tone in the can. This is the Production Checkpoint rehearsal — if it passes, Project 1's audio will too.

D. Diagnosing and prescribing

17. Fix the plan. Three problems + fixes: (1) phone across the couch = mic far away → thin, roomy audio; get a mic close (lav or boom), or at least move the phone in. (2) living room may ring / have a TV, fridge, HVAC → kill the motors, deaden if it rings. (3) no plan for room tone or a scratch/backup → grab 30–60 s of room tone; if double-system, clap to sync. Bonus: they set no levels and won't monitor — add headphones and a -12 target.

18. Fix the shoot. Sorted: Room — glass-walled conference room rings badly → get closer, hang absorption, or move to a softer room. Mic-mount — lav on a stiff nylon jacket → rustle; re-mount under a stable collar, away from synthetics, with a cable loop. Noise-source — air conditioning running → turn it off (note to restore); aim rejection at it; high-pass on. Monitoring — no headphones → they caught none of the above live; closed-back cans on, listen-back before rolling.

19. Settings drill: three rooms (model). (a) Quiet interview: average -12, set-and-forget, high-pass optional, safety track optional, tame the room. (b) Wedding speech, one take: average -14 (extra headroom for cheers/applause), safety track ON, AGC off, monitor throughout, board feed + your own mic if possible. (c) Windy rooftop: average -12, furry cover + body-block, high-pass ON, watch for gusts clipping — conservative level, ride down on strong gusts.

20. The clipping autopsy. In plain language: clipping chops the top off the sound wave when it's too loud for the recorder; that shape is now baked into the file, so lowering the level just makes the distortion quieter — there's nothing underneath to recover. On set it should have been recorded quieter, averaging -12 with headroom, testing the loudest line first. The recorder feature that would have saved an unrepeatable moment: a safety/backup track recording ~12 dB lower, which captures the loud peak clean.

21. Rescue triage. (a) Steady air-con hum → post can reasonably reduce it (steady = learnable by noise reduction, Ch.33) — but better killed on set. (b) Bathroom reverb → on set only; post can't cleanly un-echo. (c) Single clipped laugh → on set only (headroom/safety track); can't rebuild a clipped peak. (d) Wind roar → on set only; broadband chaos resists cleanup. (e) One phone buzz → post can often cut/patch it with room tone, or re-take — the least damaging because it's brief and isolated.

E. One subject, many treatments

22. Five distances. Expected: at 15–30 cm the voice is close, warm, and dominant; by 60 cm the room starts creeping in; at 1.5 m reverb and noise are competing; on the camera mic across the room the voice is thin and the room wins. The room "starts to win" typically around the point where you double past ~60 cm. Most usable: as close as you can keep the mic out of frame — usually 15–30 cm. Lesson: proximity is everything.

23. Five rooms, one line (model). Typical ranking best→worst: closet full of clothes (dead, dry, ideal) → carpeted room → outdoors (dry but with ambient hiss/wind risk) → bare room → tiled room (worst, rings hardest). "Get closer + deaden" can rescue the bare and even the tiled room substantially (proximity + absorption); it cannot fix wind outdoors (different problem — needs wind protection) and can't help much once you're already close in the dead closet.

24. Five ways to sync. Reliability/speed ranking (best first): (1) software waveform auto-sync — fast and reliable if you have a scratch track; (2) clap spikes lined up by eye — reliable, seconds; (3) timecode — most reliable at scale but needs gear; (4) a visible in-frame action (hand tap) — works as a backup clap; (5) matching lips to scratch by eye — slowest and least precise, the fallback when all else was skipped. The lesson: methods 1–2 are why you always keep a scratch track and clap.

F. Stealing techniques

25. Recreate the invisible background (guidance). Success = your subject's location "feels real" to a listener. The move: close, clean dialogue plus a generous bed of that location's own captured tone underneath (§15.6, and CS-01's Altman lesson). If the place feels fake, you probably have clean dialogue over dead silence — add the tone bed. Ask your test listener "where are we?"; a good recreation lets them describe the room.

26. Recreate a boom-op's day. Expect your arms to fail sooner than you think — often inside a minute or two at first — and your mic to drift and dip as they tire. That fatigue point is your current limit; it moves quickly with practice (and with better technique: hands apart, elbows braced, plant against your body). The lesson: booming is athletic, and consistency is the first casualty of tired arms — which is why solo shooters often clamp the boom to a stand.

G. A first taste of post

27. Sync it (guidance). With the clap, this is a ten-second job: stack the two spikes, mute the camera scratch, done — lips and clean voice locked. Note how much harder it would be without the clap (nudging by lip-reading). That contrast is the whole argument for the on-set ritual.

28. Build a tone bed. Expected: cutting two takes together, you hear the background lurch or drop at the join — a tiny hole. Laying continuous room tone underneath both makes the seam vanish; the background is now unbroken across the cut. In words: room tone is the connective tissue that makes discontinuous dialogue sound like one continuous space.

29. Patch a hole (model). Cutting out the cough leaves a beat of dead silence that reads as a glitch. Dropping a snippet of matching room tone into the gap fills it, and — because the tone matches the take's background — the deletion becomes inaudible. This is the single most common real-world use of room tone; it's why you record it every time.

30. Sound meets light (interleave). Expected conflict: the overhead boom, working into a hard key, throws a shadow on the wall/face. Resolutions that keep the close mic: work the boom from the subject's shadow side, angle the pole so its shadow falls out of frame, raise the mic slightly and re-aim, or soften the key so the shadow melts. The lesson: sound and light negotiate the same few feet above the subject — plan both together.

H. Interleaving and synthesis

31. Sound meets coverage (model). To intercut a wide and a matching medium cleanly, keep the mic position and distance consistent (or at least the tone consistent) across both sizes, keep levels identical, and — the one essential — record room tone so the editor can bed it under both and hide any small background differences at the cut. If the mic distance must change between sizes, favor keeping the lav (which rides with the subject) as the consistent spine.

32. Teach it back (model, ~200 words). "Fix it in post" is especially false for audio because the three worst location faults are all permanent at capture. Clipping: when sound is too loud for the recorder, the top of the wave is chopped off — that shape is baked in, and lowering the level just makes the distortion quieter; there's nothing underneath to restore. Reverb: echo is the voice itself, delayed and smeared into every word by the room's hard surfaces; post can't separate the voice from its own echo, so cleanup leaves it sounding underwater. Room tone: the ambient sound of a space is a real signal that only exists if you recorded it — skip it and the editor has holes of dead silence at every cut with nothing to fill them. None of these can be added later; all must be handled on set — by keeping headroom, getting close and deadening the room, and always grabbing tone. This is exactly why sound is half the picture: audiences forgive a soft image but leave over bad audio in seconds, and the half you can't fix afterward is the half you must get right on the day.

33. Lock Project 1 (guidance). The deliverable: clean, monitored, headroom-protected talking-head audio + 30 s of matched room tone, fully shot. Your three-sentence self-grade should honestly name what's clean (against the §15.3 list), what you'd re-do, and — critically — whether the room tone matches the dialogue (same mic/position/levels). Keep the note; you'll hear your own progress when you assemble the piece in Chapter 26.

I. Deeper drills

34. Settings drill: the Café order counter (model). System: double-system for isolation and safety. Mics: a lav on the customer and a lav on the barista (close, clean tracks through the noise), plus a boom favoring whoever speaks as the natural master. Levels: conservative average ~-14 dBFS with headroom, because a café is loud and unpredictable; AGC off; high-pass on to shed low machine rumble. Espresso machine: time takes around its loudest bursts, aim the shotgun's dead side at it, and accept the rest as atmosphere. Room tone: 60 s of the café's own sound after the dialogue. One-phone-one-lav constraint: sacrifice the second lav and the isolation first; protect at all costs the single close lav on the customer (the story's voice) and the room tone — a close voice over honest café ambience still reads professional.

35. Fix the shot: the lost toast (answer). The shotgun safety (a second, wired path to the voice) would have carried the six seconds the wireless dropped — redundancy is the fix. The battery habit: fresh batteries in transmitter and receiver right before the toast, never "they were fine earlier." What you tell the couple: be honest — a moment was lost, here's the clean audio you do have, and (if picture exists) whether a caption or the surrounding context can bridge it. The deeper lesson: an unrepeatable moment always deserves two independent recordings.

36. Read the sequence: boom from below (model).

FIGURE (model) — "Low two-shot, boomed from below"        [constructed teaching example]
  THE FRAME    Wide two-shot of two people at a table, low ceiling, both heads near the top of frame.
  THE MOVE     Locked off; the wide can't be cheated tighter.
  THE LIGHT    Soft, even — no hard key to throw a boom shadow (which is partly why below is safe here).
  THE SOUND    Boom from BELOW, just under the frame, angled up at the mouths. GAIN: it can get close in a
               wide where an overhead boom couldn't without dropping in. RISK: chestier tone, and the mic's
               dead side now points at the CEILING — so any vent, fan, or ceiling light ballast up there
               gets picked up. MITIGATE: check the ceiling first, kill the vent/fan, favor the talker, and
               keep a lav on each as backup.
  THE CUT      Intercuts with singles that can be boomed overhead — match the tone in the mix.
  THE EFFECT   A usable, close two-shot track where overhead was impossible.
  THE LESSON   Below is a legitimate tool with its own costs; know what your mic's dead side is now aimed at.

37. HVAC census (guidance). Expected: most "quiet" rooms have two or three continuous sources you'd stopped hearing — an HVAC vent, a fridge, a computer fan, a transformer buzz. The value is the reflex: on a real shoot you now walk in and inventory the drones before rolling, and you know which you can switch off (fridge, fans, computers) and which you can't (building HVAC — work around it with proximity and rejection). Mark the switch-off-able ones; that list is your pre-roll routine.

38. Interleave: mixed light meets clean sound (model). The conflict lives in the few feet above the subject's head. Leaving the warm counter lamps on (Ch.13's motivated practicals) is a lighting win, but if one is a hard-ish source, an overhead boom swung over the counter can throw a boom shadow across the subject or the back wall. Resolution that sacrifices neither: work the boom from the subject's shadow side and angle the pole so its shadow falls out of frame — keeping the close overhead mic and the warm practicals. If the geometry is impossible, drop to a lav (no boom, no shadow) and keep the lamps. Sound and light negotiate the same air; plan both together.

39. Build your own checklist (guidance). A strong personal list is shorter and more specific than the generic card — it names the two or three things you actually forget (e.g., "arm record on BOTH devices," "fresh wireless batteries," "kill the studio fridge," "grab room tone before packing"). If you shoot events, it'll weight redundancy and safety tracks; if you shoot solo interviews, it'll weight room-taming and the listen-back. The test of a good checklist is that it catches your recurring mistake — compare it honestly to the key-takeaways card and keep whichever line has saved you.

Chapter 16 — Answers to Selected Exercises

Model answers and critiques for the starred and odd-numbered exercises. These are models, not the only correct responses — a defensible plan that reasons well is the goal.

A. Reading the plan behind the video

1. Spot the treatment. The tell of a logline vs. a topic: a logline names a tension or a "so what" ("the last repair shop on a street of chains"), while a topic just names a subject ("a repair shop"). If your one-sentence guess could headline a hundred different videos, you've written a topic — push it until it commits to this story. The driving adjective usually reveals itself in the pace and light (a "gentle" film holds shots and uses soft light; an "urgent" one cuts fast and hard).

3. Reverse-engineer the shot list. Expect to count many more shots than you remember watching — a polished 60–90 second segment often hides 15–30 distinct shots, most of them B-roll and cutaways. The lesson: the shoot had to gather far more coverage than the runtime suggests. A finished minute is built from several minutes of captured pieces; that ratio is what your own shot list must plan for (this previews Ch.20's coverage mindset).

5. The missing plan. Model diagnosis of a chaotic video: aimless/no through-line → missing treatment (no logline; nobody decided what it's about). Shaky, redundant, or gap-filled coverage → missing shot list (no plan for what to capture; no priorities). Blown light or wrong time of day → missing schedule/recce (not built around fixed points). Echoey or noisy audio → missing recce sound check. The exercise proves the chapter in reverse: nearly every "bad video" symptom traces to an absent pre-production document.

B. From idea to plan

7. The six-question treatment. A strong model (barber shop): Logline — "In a chair that's held three generations of the neighborhood's heads, a barber gives more than haircuts." Audience — locals and lovers of small-craft docs. Format — 90 sec, 16:9, web. Tone — warm, unhurried, a little funny. So what — the viewer wants to book with a real barber, not a chain. Look — window light, close on the hands and the mirror, natural shop sound. First image: the striped pole starting to turn at open. Last image: the cape snapped off, the customer's satisfied nod in the mirror. Grade up any version that includes a genuine first and last image and a viewer action.

9. The client swap test. If a colleague's treatment for a bakery reads fine when you paste in "hardware store," it's a topic. The fix is specificity that could only be this subject: the ninety-year-old sourdough starter, the 4 a.m. start, the regular who's come daily for thirty years. What you had to add to break the swap — concrete, unique detail and a specific first/last image — is exactly what turns a topic into a plan.

C. Scripting

11. Write a cold open. Grade the AUDIO column hardest: every row must have sound planned (room tone, VO, music, SFX), or the student is forgetting sound is half the picture. A strong open pairs an arresting first image (a detail, not a wide) with a hook line of VO, establishes the place by row 2–3, and names the music entrance. Bonus for marking the opening lines as VO (recordable at the interview, laid over B-roll) rather than sync.

13. Two-column vs outline. (a) Scripted product promo with narration → two-column (words written in advance). (b) 5-minute teacher profile → outline (the teacher's words can't be scripted) + a two-column open/narration if any. (c) How-to explainer you narrate → two-column. (d) Wedding highlight film → mostly outline/shot list (no script; it's found moments cut to music) — arguably neither classic form, planned as a shot list and a music-driven structure. The reasoning matters more than the label.

D. Boarding and listing

16. The twelve-row shot list. A strong list: uses the full column set (size, move, subject, audio, priority); has ≥6 coverage/B-roll rows against the hero shots; marks every row 1/2/3; includes a room-tone row (priority 1); and encodes at least one cross-chapter reminder (light, the line, a release). Dock any list that is all hero shots or has everything marked "must-have."

17. Prioritize under fire. The point: if the student can't cut their list roughly in half and still have a cuttable film, their priorities are soft. A healthy Project-2 list has maybe 5–8 priority-1 rows (open, interview, hero B-roll, one workday wide, close, room tone) and the rest as 2/3. "Everything is essential" is the failure mode; force a ranking.

18. When NOT to storyboard. Board: staged action, a camera move that must hit a mark, VFX/composites, a precise commercial product shot, a designed open/reveal. Don't board: a plain talking-head interview, found/observational B-roll, run-and-gun event coverage. The rule: board what is designed and can only be got once; list what is found. Storyboards plan invention; shot lists ensure coverage.

19. Board the line. A correct answer marks each person's screen position across the four frames and confirms the interviewer/subject keep consistent left/right placement so eyelines match. A crossed line shows up as two people who appear to look the same direction (both screen-left) instead of at each other — the board catches it because you can see both frames side by side before the shoot. Fix on paper by keeping the camera on one side of the line (Ch.9).

20. Coverage audit. Any interview or hero setup with no cutaway is a trap: in the edit you'll want to shorten it and have nothing to cover the trim, forcing a jump cut (Ch.20's logic). The fix is to add at least one cutaway per setup — a detail, a reaction, a wide — even if it feels redundant on set. Redundant on the day, essential in the timeline.

E. Scheduling

21. Story order vs shooting order. Model: shoot the interview first (quiet, controllable, fragile), then B-roll in the same location/light while set, then the wides/exteriors grouped by where they are and when the light is right — even though the video plays wide → interview → B-roll → wide. Each reordering reason: group by location/light, capture fragile shots first, avoid company moves.

23. Estimate honestly, then add a third. A realistic lit interview: lighting ~20–30 min, settling the subject ~10–15 min, shooting coverage ~20–30 min → ~50–75 min, plus a third of buffer → budget ~70–100 min. The beginner's optimistic "we'll knock out the interview in 20 minutes" breaks at the first hitch (a nervous subject, a hunting autofocus, a passing truck) and cascades into the rest of the day. The buffer is what absorbs the inevitable.

25. Three currencies. Model (Project 2): Money — ~$25 food + a drive. Time — a full Saturday plus 4 hours planning. Favor — the subject's hour and a friend on the boom; repay by screening the film for them and crediting them, and by boom-op'ing their next shoot. Grade for naming a concrete repayment, not just the favor.

27. Rent, borrow, or buy. Tripod → buy (used constantly). Cinema long lens for one shoot → rent (specialty, occasional). Lav mic → buy if you shoot dialogue often, else borrow. Gimbal used twice a year → rent or borrow (rare use). Spare memory card → buy (cheap, always needed). Rule: buy the daily items, rent the specialty item, borrow the rare one.

29. The sound scout. Students almost always play back sounds they never consciously heard: an HVAC hum, a fridge compressor, distant traffic, a buzzing light, a neighbor's music, room echo. That's the entire point — the mic hears what the distracted ear filters out, so you must listen on headphones on the recce. If it's bad, you fix it now (kill the source, add soft furnishings, or change rooms), because you can't un-echo a room in post (Ch.15).

F & G. Budget / Scout

22. Build the day. A strong schedule: grouped by setup/location; built around at least one fixed point (dawn light or the quiet hour) with that fragile shot captured first; a buffer block; must-haves front-loaded. Dock schedules in story order or with zero slack.

24. Two fixed points, one day. Sequence: golden-hour exterior at the top (dawn), quiet interview next (before crowds/noise), controllable B-roll in the flat mid-day hours (light quality matters less for detail/cutaway work, or use open shade), then return for the dusk exterior. The trick is scheduling the two light-dependent, unmovable shots at the ends of the day and filling the "ugly light" middle with work that doesn't depend on beautiful light.

26. The zero-dollar budget. Grade for honesty: money column near zero only because costs moved to time/favors, AND the food line, contingency, and "your own time" line are present. A budget that reads "$0, done" while hiding a full weekend and three unrepaid favors has failed the exercise's real lesson.

28. From budget to bid. Convert by adding a realistic day rate for your time (the biggest line a passion project absorbs for free), marking up rentals, and adding a margin/contingency. The single quoted number should make explicit which lines the client now pays for (your time, insurance, a real music license, revisions) that the passion project ate. Previews Ch.38.

30. The five-domain recce. A complete answer fills all five domains with specifics (not "sound: fine" but "sound: fridge hum from the back; quiet before 9"), includes three reference photos, notes light direction and time, and — crucially — was done at the shoot's time of day. The most common gap is a missing or vague sound entry.

31. Draw the recce map. A strong map places subject, camera, light source (with direction arrow), mic, and any bounce using the frozen symbols, plus notes on power, sound, and permissions. It should translate the scout into a shootable setup — someone could walk in and build it. See FIGURE 16.7 and CS-02's FIGURE CS2.2 as models.

32. Plan B. Grade for a concrete backup tied to the biggest surfaced risk: "if the workday is too loud, move the interview to Sunday dawn"; "if it rains, the exterior moves a week and we shoot the interview under the porch." A named alternative — different room, time, or approach — not just "we'll figure it out."

H. Fixing broken plans

33. Fix the shoot day. Problems: (sound) interview in a café at 10 a.m. = peak noise; (light) exteriors at 11 a.m.–noon = harsh overhead sun; (order) bouncing café→outside→café wastes moves; (buffer) no slack. Rewrite: shoot the interview first, before the café opens or in its quiet hour; exteriors at golden hour (early or late), not noon; group the café shots together; add buffer. Same shots, calm day.

35. Fix the treatment. The original makes no decisions. A fixed version invents specifics and commits: a logline naming a real story ("the founder still answers the support line herself"), an audience, a length, a tone, a first and last image, and a viewer action. Explain that you added decisions — the whole function of a treatment is to decide, and the original decided nothing.

I. Interleaving

36. Coverage on the page (Ch.7). To shorten one interview answer without a jump: put on the list at least a medium (the answer lives here), a wide/2-shot (cutaway for scale or to hide a trim), a close-up (to punch in on the key line), and at least one B-roll cutaway of what they're describing. Each lets you cut away and back, hiding the trim. Minimum viable: the answer size + one cutaway.

37. Continuity enables the schedule (Ch.9). Shooting out of story order is safe only because continuity — matching action, consistent screen direction, staying on one side of the 180° line — lets the pieces reassemble into a seamless flow in the edit. You must shoot carefully: match the action across sizes, keep eyelines and screen direction consistent, and note prop/wardrobe/light continuity, so a shot captured at 7:45 cuts invisibly against one captured at 9:00.

38. Directing in pre (Ch.10). Three pre-production decisions for a nervous subject: (1) schedule buffer so they can warm up (don't roll on the first stiff take); (2) plan a minimal, unintimidating setup (small crew, gear tucked back) on the recce; (3) pre-write open, conversational questions and decide an eyeline/blocking so they're never guessing where to look. The treatment doubles as reassurance — show them what the film is.

39. Light and sound are pre-production decisions (Ch.13, 15). Students should mark: on the schedule, the golden-hour and quiet-hour blocks (light + sound decisions); on the recce map, the window key and the "kill the fridge" note (light + sound); on the shot list, the "dawn light only," "2-mic," and "room tone" rows. The argument writes itself — most of "fix it in pre" is deciding, in advance, where the light and the quiet will come from.

J. Project

41. Complete the Production Checkpoint — rubric. A strong Project-2 pre-production package: - Treatment: answers all six questions; has a genuine first and last image; commits to a tone and a viewer action; would survive the swap test (couldn't be pasted onto a different subject). - Shot list: 12+ rows in the FIGURE 16.4 format; ≥ half coverage/B-roll; every row prioritized; the priority-1 rows alone would yield a cuttable film; includes a room-tone row. - Schedule: efficient (grouped) order, not story order; built around at least one fixed point (light or quiet); fragile/must-have shots front-loaded; a buffer block present. - Cross-chapter integration (bonus): the plan visibly applies coverage (Ch.7), continuity (Ch.9), light (Ch.13), and sound (Ch.14–15).

Keep all three documents together — this is the folder the student shoots Project 2 from in Chapters 19–20.

K. Deeper drills

42. Settings Drill: from tone to settings. Model (tone = "patient, unhurried"): support → locked-off tripod (stillness reads as calm); shot length → held, longer shots in the edit; light → soft (window/diffused), low contrast; sound → quiet, natural, room to breathe, minimal music. Biggest pre-production decision that follows: schedule the shoot for the soft-light hours (dawn/dusk or overcast) — the tone requires a light that only the schedule can guarantee. Flip the adjective to "urgent/gritty" and every row inverts (handheld, quick cuts, hard light, driving sound).

43. Five Ways: one subject, five loglines. Grade for five genuinely different films, not five rewordings — each should imply a different audience and structure. The strongest choice for a 3-minute doc is usually the one with a clear human tension and a "so what" a stranger can feel in ten seconds (e.g., the day-in-the-life of a librarian who quietly runs the block's only warm public room, over the "buildings-as-democracy" essay, which is harder to make land visually in three minutes). Reasoning about hold-ability matters more than the pick.

44. Recreate the plan. A strong answer shows the polished piece was planned: a plausible logline; the headline shots that clearly had to be on a list (an establishing shot, a hero shot, specific cutaways, a bookend); the scheduling logic (shot at golden hour? in one location? around a real event or crowd?); and the budget currencies (money on gear/location/talent; time on the shoot; favors where visible). The point students should reach: nothing that looks effortless was unplanned — the effortlessness is the planning.

45. Read the plan: spot the gap. The most conspicuous absence is a cutaway/B-roll of what the subject is talking about (and/or a second interview size). With only one interview size and a couple of generic shots, there's nothing to cover the trims when you shorten the answer — every cut inside the interview jumps. Fix: add B-roll matched to what's said, plus a close-up size, so you can cut away and back invisibly (Ch.20 logic).

47. Settings Drill: the recce-to-settings bridge. The findings dictate: when — shoot the interview 7–8 a.m. (soft east light) and before the 9 a.m. public opening (quiet), never mid-morning; light — use the window as a soft key, possibly a bounce for fill, and avoid the hard post-8 sun (diffuse or reposition); audio — kill the fridge (unplug it for takes) and shoot the interview in the quiet pre-open window, two mics for safety; power — no outlets means everything runs on charged batteries, so charge overnight and bring spares. Every one of those is a settings decision made on the recce, in pre — the essence of the chapter.

Chapter 17 — Answers to Selected Exercises

1 (Find the two act breaks). Grade on whether the student locates a real inciting incident (the moment the normal is disturbed and the story starts) and a real turn toward resolution, not arbitrary timecodes. Strong answers state the arc as a change: "at the start he's about to quit; at the end he's found a reason to stay." If they can't find an arc, that video may genuinely lack one — a valid finding.

3 (The open-loop map). A good map names specific questions ("will the recipe survive? will the shop sell? does the experiment work?") and shows at least one loop opening early and closing late. If a video held them with no open loops, push them to look again — often the loop is implicit ("where is this going?") or the hold is doing something else (pure spectacle, comfort, familiarity), which is worth naming as the exception that proves the rule.

5 (Frame Log — the story week). No single right answer; reward consistency and specificity. The payoff is the meta-observation at week's end — typically "hooks are everywhere and most are the first spoken line," or "the videos that dragged all had long stretches where nothing turned." That noticing is the eye this chapter trains.

6 (Name the beats — the search box). Beat 1 = hook + setup (a life decision in five words); beats 2 = setup→development (arrival, exploration); beat 3 = complication (separation); beat 4 = turn (he commits/moves); beat 5 = resolution begins (wedding); beat 6 = resolution lands (a child). The hook works in five words because it opens the biggest possible question — whose life is this, and what will happen? — while promising a whole story will unfold.

7 (Diagnose the beat map — bus driver). Example, Act 1: A-beat — "he's driven route 12 for 25 years, 4:50 a.m. start." B-beat — "he knows every kid's name and which ones had a hard morning." The A informs (the job); the B moves (what the job means). Good answers keep the A concrete and the B human, and make them the same moment seen two ways.

8 (The missing turn). It's a list because every beat connects with "and then" — nothing causes the next, nothing is at stake, and nothing changes meaningfully. Rewrite with a turn and causation: "She opened the bakery, but a chain moved in across the street, therefore she bet everything on a 4 a.m. sourdough nobody else made, until finally the line out the door was for the one thing the chain couldn't copy." Now there's a complication, a decision, and a change.

9 (Critique a real arc — Up). Act 1: the wedding and the two personalities established (setup); the inciting incident is the shared dream of Paradise Falls (the goal that will be deferred). Act 2: the jar filled and broken, life intervening, the loss of the hoped-for child — rising complications. Climax/turn: Ellie's death. Act 3: Carl alone, changed. Logline, e.g.: "A tidy, cautious man and his bold wife save for the adventure of a lifetime that ordinary life keeps deferring — until she's gone and the dream is his alone to carry." Reward any version that finds a genuine arc of change.

10 (The one-line logline). Strong loglines are specific and contain a change or a stake: "A retired machinist rebuilds the town's broken clock tower to prove the trade he gave his life to still matters." Weak ones are topics, not stories: "A video about a clock tower." If the student can't write the sentence, that's the useful finding — the story isn't found yet.

11 (Fill the spine). Most beginners leave the "Until one day…" blank empty or weak, because they forget a story needs a turn — an event that disturbs the normal. The fix is to hunt for the real inciting incident in the subject's true history (the diagnosis, the closure, the decision, the loss). A spine with a strong "Until one day" almost always structures itself from there.

12 (But / therefore). Grade on whether the rewrite genuinely forces causation. "I took a job, and then I moved, and then I quit" becomes "I took a job across the country, but I hated the work, therefore I used the savings to start the thing I actually wanted." The lesson: "and then" hides the story; "but/therefore" reveals it.

13 (Three acts for a testimonial — bike shop). Model: Hook/Act 1 — "My commuter bike died the morning of a job interview." Act 2 — "Every shop said three days. This one said, 'sit down, we'll have you rolling in twenty.'" Act 3 — "I got the job. I've bought every bike here since." A change (stranger → loyal customer), a real stake (the interview), and a hook that isn't a résumé line.

14 (Two structures, one subject). Neither is wrong; the person-arc is usually stronger for emotional engagement (viewers bond to a face and a change in a person), while the place-arc can be stronger for scope or theme. The trade-off: a person gives you a clear B-story and a face to follow; a place gives you breadth but risks having no one to root for. Best answers pick one and justify it by what the video is for.

15 (The reality-arc pivot). Re-outline around the succession story: hook on a quiet moment of the nephew getting something wrong and the owner's patience; Act 2, the stakes (the shop's future, a family legacy); Act 3, a first solo success. You'd need to shoot: teaching moments (hands-on-hands), the nephew alone, the owner watching, and an interview about why passing it on matters — none of which the reopening plan called for. The lesson: shoot enough coverage that a truer story, if it appears, is filmable.

16 (Delete the throat-clearing). Rewrite, cold: "I flooded my kitchen twice before I learned the one thing every plumber knows about a leaky faucet." The subscribe ask and channel branding move to the end, after you've earned the viewer. The first line now opens a loop (what's the one thing?) instead of clearing a throat.

17 (Five hooks, one video). For, say, a video about a 90-year-old marathoner: Question — "What would you do with an extra forty years of running?" In medias res — open mid-race, at mile 20, on the face. Striking image — a wall of forty finisher medals. Bold claim — "He's run a marathon every year since before you were born." Promise — "Here's what nine decades taught him about not quitting." The bold claim or the image usually wins for a stranger, because both create instant curiosity without requiring words.

18 (The buried hook). It's normal because people warm up as they talk — the electric, distilled sentence usually arrives once they've relaxed and found the heart of it, well into the interview. In the edit you lift that sentence and lead with it, then fill in the setup afterward. This is "shoot for the edit": you couldn't script which sentence it would be, so you recorded generously and found it in post.

19 (Recreate It — the search-story). Grade on structure, not polish: six beats, a five-word hook that opens a life question, causal middle, and a resolution that lands with implication (an object or search that says "and everything changed") rather than statement. The constraint (no dialogue) forces pure show-don't-tell — that's the skill being built.

20 (Adjective to evidence). Examples: generous → he waves off payment from the kid counting coins. Exhausted → she sits in the car for a full minute before going in. Brand-new → the plastic still on the seat, the sticker on the window. Beloved by regulars → three people say "the usual?" and she already knows. Meticulous → he redoes a seam nobody would ever see. Any concrete, observable answer beats the adjective.

21 (Cut the redundant narration). The problem: the VO names exactly what the picture already shows, insulting the viewer and wasting the audio track. Rewrite the VO to carry what the image can't: "She's made this dough the same way for thirty years — the recipe is the only thing in the shop she's never changed." Now word and image do different jobs.

22 (The feeling and the proof). Three proofs for "thirty years and I still love it": (a) the small private smile when a piece comes out right; (b) hands that move without looking, worn smooth into the tools; (c) staying late, unpaid, to redo one he isn't happy with. The third is strongest — sacrifice is the most convincing proof of love, because it costs something.

23 (Shoot This — a trait without words). Self-critique target: name whether your viewer got the trait unprompted, and if not, which shot was too vague. Strong trait-shoots pick specific behavior (the redone seam, the coins waved off) over generic activity (someone just working). The fix for a miss is always a sharper, more particular piece of evidence.

24 (Fix the flat middle). Cause 1: the middle doesn't complicate — it adds facts that don't raise stakes or doubt; fix by finding the real obstacle and building the middle around it. Cause 2: the beats are too far apart — long stretches with no turn; fix by cutting dead material so turns arrive more often. A sagging middle is almost always "no rising complication" or "beats too sparse."

25 (Fix the all-middle video). Missing Act 1: no setup that makes us care — add an opening beat that names a stake or a person ("this 40-year-old machine runs the whole line"). Missing Act 3: no resolution — add a closing beat that lands a change or a result ("and it just shipped its millionth part"). Bookends turn a pile of information into a story.

26 (Fix the spoiled ending). Front-loading "how we saved the shop" hands over the ending, killing every open loop — there's nothing left to discover, only reasons to sit through. Rewrite to open on the threat: "The bank gave us thirty days." Now "we saved it" becomes a payoff the viewer travels toward instead of a spoiler they start with.

27 (Structure drill). (a) Product ad — hook: bold claim or striking image; arc: "a nagging problem → this solves it → a better everyday." (b) Nonprofit profile — hook: character or in medias res; arc: "a person in need → the program's turn → a changed life." (c) How-to — hook: a promise; arc: "you can't do X → the steps → now you can." Reward sound reasoning over "correct" labels.

28 (Cut This — reorder to the arc). Guidance: the shot order almost never survives contact with the arc. The re-ordered cut typically wins by leading with the strongest beat and ending on a change; the shot-order cut typically buries its best moment and stops when footage runs out. The learning is felt, not argued — students should watch both and notice the pull.

29 (Cut This — find the hook in an interview). Grade on whether the 20–30s cut genuinely opens on the strongest line and then back-fills setup without confusing the viewer. This is the in-medias-res move at micro scale and a direct preview of the paper edit (Ch.29–30). Even a rough version teaches the core lesson: the best line usually belongs first.

30 (Story meets shot list). Guidance: each beat should generate at least one interview/action shot and one show-don't-tell B-roll shot. A good deliverable reads like a plan — "Beat 3 (the obstacle): interview line about the setback + B-roll of the empty order board." This is the exact bridge from Chapter 16's shot list to a story that will actually cut.

31 (Story meets coverage). Model: for a "the obstacle" beat, shoot a wide (the empty shop for context), a medium (the owner's posture), a close-up (the worried hands or face), and an insert (the overdue notice). The edit can then build the beat's mini-arc — establish, feel, land — because you gave it the pieces. Tie every shot to the beat's job.

32 (Treatment from spine). A strong treatment paragraph states the concept, the tone, the arc, and the hook in readable prose a client could approve: "A three-minute portrait of the last cobbler on Main Street. Warm, observational, lightly scored. We open on his hands resoling a boot at dawn and follow one working day that becomes the story of a vanishing trade — and one man's refusal to let it vanish quietly." Grade on whether a stranger could greenlight it from the paragraph alone.

33 (Project 2 — draft the arc). Rubric: (1) a real logline with a change/stake; (2) distinct A- and B-stories; (3) 5–8 beats in three acts, ordered for the arc, connected by but/therefore; (4) a marked hook that is the strongest beat, up front; (5) a genuine change in the final beat, not a logo; (6) one show-don't-tell shot per beat. Dock for chronology-order outlines, missing turns, weak hooks, and endings that stop rather than resolve. This card is the spine the rest of Project 2 is built on — grade it as the load-bearing document it is.

34 (Teach it back). A strong ~200-word explanation shows that a testimonial holds the same way any story does: it opens a loop (a problem), keeps the outcome in doubt, and pays it off with a change (problem → trust). Grade on whether the example actually demonstrates the arc (the flooded basement → one plumber answered → never called anyone else) rather than just asserting "it's a story." Clarity for a true beginner is the target.

35 (The structure autopsy). No right scores; the value is an honest, specific diagnosis. Strong autopsies name a fixable weakness ("no arc — nothing changes; I'd add a stake in the first ten seconds and a result at the end") rather than a vague one ("it was boring"). Whichever axis scores lowest points to the section of this chapter to re-read. Re-grading after Part VI measures real growth.

36 (Same ore, three arcs). The exercise proves that shape is a choice but not an arbitrary one — the honest fit is the shape the material's true change already makes. The bike-shop ore fits "man in a hole" (nearly lost it → rebuilt it, stronger) most naturally; forcing it into "rags to riches" would overstate the triumph, and "quest" would overstate a single goal. A forced fit shows up as beats you have to invent or exaggerate; a true fit uses only beats that actually happened.

37 (Three scales). Grade on consistency across scales: the logline, the treatment paragraph, and the six-word poster line should all be about the same arc and stake. If the treatment introduces a story the logline didn't promise, or the poster line can't capture the change, the story isn't settled yet. Example poster line for the cobbler: "The last man who fixes things." The scaling exercise is a clarity test, not a copywriting one.

Chapter 18 — Answers to Selected Exercises

Model answers and critiques for the starred and odd-numbered exercises (plus the even ones flagged "provided"). These are models, not the only right responses — a defensible answer that reasons from the chapter's principles is the goal.

A. Seeing the set

1. Spot the crew. Visual tells to look for: the operator has the camera and is watching the frame; the DP is often near the monitor, not the camera, and is looking at the look; the AC is beside the camera with a hand near the lens (pulling focus) or holding a slate/managing cards; the gaffer and grips are near the lights and stands (gaffer at the units and power; grips at the rigging, flags, and sandbags); sound is the one in headphones, often with a boom or a mixer bag; PAs are at the edges — locking up, wrangling cable, holding umbrellas, carrying things. The lesson: a set is a division of labor you can read off people's hands. If you can't tell who owns what, it's usually a badly run set.

3. The one-take moment. Model: an award moment or a first kiss is caught only because the camera was framed, exposed, focused, and rolling before it happened, and someone had audio ready for it. What had to be true in advance: the operator knew where to point (someone had the run-of-show), exposure was set for that spot, and the record button was already down. Connect to FIGURE 18.8: "the organization existed before the moment did." A crew that reacts to the moment has already lost it — you cannot roll, frame, and expose for a three-second event after it starts.

B. Reading the day

5. Read the caught shot. The four things ready before the moment in FIGURE 18.8, mapped to roles: (1) support + frame — the operator was already framed and on a stable rig → camera/grip; (2) exposure locked earlier for that spot → DP; (3) sound fed clean from the board + camera mic ready → sound; (4) the cue to roll early ("award in two minutes") → the PA watching the run-of-show. The moment was caught because four roles had each done their job before the moment, not during it.

6. Fix the call sheet. "Shoot at Dave's house, Saturday, bring cameras" is missing at least: a specific call time (people drift in all morning; nothing starts) → wasted hours; the address + parking → late, lost crew; a schedule (what's shot when) → chaos and arguments on the day; a meal time → a hungry, slow, cranky crew; the nearest hospital → panic if someone's hurt; a must-get list → nobody knows what to protect if time runs short; contacts → no way to reach a missing person; a wrap time → the day sprawls to midnight. Each omission maps to a §18.2 field and a predictable failure.

7. Fix the etiquette. Three breaches: (a) the crew member walked through the back of the frame during a take — they should have frozen at "action" and waited for "cut" (quiet + stillness, §18.4); (b) the phone buzzed — phones off/silent before rolling; (c) the operator saw a leaning, unweighted light stand and did nothing because "not my department" — safety overrides "stay in your lane"; anyone who sees a hazard alerts the responsible person or calls it out immediately. The correct principle: stay in your lane for creative decisions, but never for safety.

8. Triage this morning. Order (impact, not arrival): (a) the cable trip hazard first — safety outranks everything; someone could fall (clear/tape it now). (b) the AC-hum in the interview corner second — audio is hardest to fix and loses viewers (kill/mask the source before rolling). (c) the dim key light third — a stop is usually recoverable in post, and it doesn't threaten a person or the dialogue. (d) the forgotten spare battery last — it's a redundancy gap to note and manage (ration/charge), not an active fire, unless it actually dies. Justify each in one line as above. (FIGURE 18.7.)

9. Diagnose the disaster. Post-mortem, ranked by damage: - Reused the only card twice when it filled → overwritten, permanently lost footage. Worst possible; §18.3 (never erase until backed up in two places). Irreversible. - No slates + identical filenames → the edit can't find anything; a day lost to archaeology. §18.3. - No shot list / no plan for what to capture → probably missing the shots the piece needs, with no way to know. §18.2 / Ch.16. - One battery → the shoot could have died at any moment; pure luck it didn't. §18.6 redundancy. - No call sheet → no schedule, no must-get, no hospital; the midnight wrap is the symptom. §18.2. - Wrapped at midnight → fatigue/safety risk driving home. §18.4. The ranking teaches the lesson: the unfixable failures (lost footage, missing shots) outrank the merely painful ones (a long day).

C. Building the documents

10. Fill a slate. Model completed slate: PROD: Hands of the Trade | ROLL: A001 | SCENE: 3 (interview) | TAKE: 2 | DATE: 12 Oct | FPS: 24 | DIR: (you) | DP: A. | ☑ SYNC. If shooting MOS (e.g., silent B-roll), check MOS instead. The grade is whether every field is filled and whether the student would actually read "scene three, take two" aloud — the spoken slate is what saves a soft or off-frame board.

11. Write the call sheet. A strong model has every field of FIGURE 18.2 filled for a real half-day: production/date/day, crew call + staggered talent call, est. wrap, location + address + parking, weather + sunrise/sunset, nearest hospital + route, full contacts, an hour-by-hour schedule with a meal and a buffer, and a must-get list of 2–3 items. Grade hardest on the fields beginners drop: hospital, staggered call times, meal, wrap, must-get. A call sheet missing the must-get list has missed the point — it's the one field that saves a compressed day.

13. Stagger the call times. Model around a 09:30 first shot: Producer + DP + sound: 08:00 (load in, set up, light, frame, sound-check — setup always runs long); subject: 09:15 (welcomed, mic'd, relaxed just before we roll — not left waiting an hour, which kills the performance); first shot: 09:30. Justify: crew early because setup is the underestimated part; talent late enough not to sit around getting anxious (a Ch.10 performance concern) but early enough to settle. The staggering is the craft — it respects both the clock and the human on camera.

D. Speaking and running the set

15. Say the calls. Canonical order: "Quiet on set" → "Roll sound" → (Sound) "Speed" → "Roll camera" → (Camera) "Rolling" → slate + clap ("scene , take ") → "Action" → [take] → "Cut" → "Check the gate" / "Going again" or "Moving on." End of the whole shoot only: "That's a wrap." Grade: did the student put sound before camera (sound rolls first so it captures the clap) and slate after both are rolling? Those two orderings are the ones people get wrong.

17. Recreate a role handoff. Most students report that sound is harder to own with full attention than picture — because picture problems are visible (you see soft focus, bad framing) while audio problems are only audible and easy to tune out while you're busy. That's exactly why sound is a separate role: the person on picture literally cannot also be listening properly. The handoff teaches that "one person, all jobs" isn't heroic — it's how things get dropped.

E. On the day

19. Shoot to a call sheet. Guidance: the value is in the divergence log. Students almost always find setup ran long and they got fewer shots than planned — which is the §18.2 lesson (double your setup estimate; front-load must-gets) learned in their own hands. A strong reflection names where reality diverged and what they'd change on the next sheet (more buffer, earlier call, fewer optional shots).

21. Two-person shoot, sound owned. Model finding: with a dedicated set of ears, the audio is dramatically cleaner — the sound person catches the mic that came unclipped, the AC hum, the plane overhead, the level that's too hot — all things a solo operator misses because they're framing. The reflection should conclude that sound is the highest-value second person precisely because it's the department a solo shooter can't monitor. (This is the §18.5 rule, proven personally.)

F. Organizing for the edit

23. Log ten takes. Model log row: Scene 3 / int. / take 2 — GOOD ✓ (circled) — "full-sentence answer, warm, looks at lens". A strong log has: an address (scene/shot), a take number, a one-word verdict, a circle on keepers, and — the pro move — why it's the keeper. The "why" is what future-you trusts. Grade down logs that only record take numbers with no verdict; that's a list, not a log.

25. The findability test. Expected result: finding "take 3, the circled one" in a slated/logged folder takes seconds; finding it in an unlabeled folder takes minutes of scrubbing — and that gap, multiplied across a whole edit, is hours to days. The proof of §18.3: organization isn't neatness for its own sake; it's time, and the time is spent on set (cheap) or in the edit (expensive). The clap and the spoken slate are the index to your own footage.

G. Judgment and safety

26. Five ways to staff one shoot. Model (talking-head interview), showing which job gets its own body first as the crew grows: - 1 person: you do everything, in passes (§18.5). Sound is the risk. - 2 people: split picture from sound — the highest-value first split, because sound is what a solo shooter drops. - 3 people: add someone on light + support (a gaffer/grip hat), so the DP can concentrate on frame and the subject. - 5 people: now the roles separate — director (performance) distinct from DP (image), dedicated sound, a gaffer, and a PA (lock-up, wrangling, the subject's comfort). - Full small pro crew: each department is its own person — director, DP, operator, AC (focus/media), gaffer, grip, sound, PA(s), producer. The pattern: the first thing you buy with a second body is ears; the next is light; the next is the split between story and image.

27. The safety call. Three "stop the shoot" situations and who may call it: (a) a light stand tipping toward a person — anyone; (b) a cable across a wet floor / overloaded power — anyone, especially the gaffer; (c) someone about to do something dangerous to get a shot (climb an unsecured height, step into traffic) — anyone, and especially the person being asked to do it. The unifying rule (§18.4): anyone can and must call cut for a genuine hazard, and a good set thanks them.

28. No shot is worth a person. Model decline: "That's not a shot we're going to get that way — it's not safe, and no shot is worth someone getting hurt. Here's how we get the same story safely instead." Two safe alternatives (e.g., for a "in-traffic" hero shot): (1) shoot long from a safe position so the subject is nowhere near the danger and the lens compresses the distance; (2) stage it in a controlled/closed space or use a safe foreground element to imply the danger. The graded point: you decline and solve the story problem — safety and craft are not opposed; the constraint forces a better, safer idea.

H. Interleaved

29. The shoot that applies everything. Model one-line-per-chapter plan for a one-hour talking-head-plus-B-roll shoot: Exposure (Ch.5): set ISO/aperture with the waveform and lock it. Shots/coverage (Ch.7): a wide + matching medium/CU so it cuts. Continuity (Ch.9): matching action on any B-roll so inserts fit. Directing (Ch.10): an easy warm-up question + eyeline just off-lens for a relaxed subject. Light (Ch.11): window key + a bounce fill. Sound (Ch.15): mic close, room tone recorded, monitored on headphones. All of it run with a call sheet, the seven passes, and a slate on every take. The grade is whether each earlier skill is actually used, not just named — and whether the day is run (sheet/passes/slate), which is the Ch.18 contribution.

30. Pre-to-set handoff. Model: turning the Ch.16 treatment/shot list/schedule + Ch.17 arc into a call sheet reveals what planning leaves out and set-running adds — chiefly the logistics and safety layer: individual call times, the location address and parking, contacts, weather/light, the nearest hospital, the meal, and the must-get distilled from the shot list's priorities. The insight to grade for: a call sheet is not the schedule retyped — it's the schedule plus everything a human needs to physically show up, stay safe, and know what to protect. The handoff from plan to set is where a project stops being an idea and becomes a day.

Chapter 19 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered items. Interview craft is judged on effect, not on hitting one "right" answer — use these as calibration, not gospel.

A. Seeing and hearing the craft

1. Find the eyeline. Most documentary and news interviews use an off-axis eyeline (subject looks just off the lens at an interviewer) — it reads as "we're watching a real conversation." Direct-to-lens is rarer and reads as address (a host, a spokesperson, or a deliberate Morris-style choice). If you found one direct-to-lens interview, ask what it was going for — was the viewer being implicated or sold to?

3. Where's the key? Model observation: in a well-lit talking head the key is usually on the side the subject faces (the looking side), soft, just above eye level, with the shadow falling on the camera-near cheek. A bright, even face (near-1:1 fill) signals corporate/testimonial; a deep shadow (4:1+) signals a moodier, more serious documentary tone. If you can't tell where the key is, that's often the mark of flat, unmotivated lighting — a finding in itself.

4. Hear the interviewer. In a clean interview edit you hear only the subject — the questions are cut out (§19.5's whole point), and there are no "mm-hmm"s bleeding over the answers. A clean interview edit sounds like a monologue assembled from many answers, with continuous room tone and no verbal clutter from the interviewer. If you can hear the interviewer's "right, right," that's the exact mistake §19.4 warns against.

5. The non-answer hunt (model). Example logged: Interviewer asks "So that was a tough year, huh?" → subject: "Yeah, definitely." That's a non-answer (a leading question producing a yes/echo). Rewrite: "Tell me about that year — what made it hard?" Two more common patterns: a fragment ("Since 2015.") rewritten by briefing the full-sentence rule; an evasion ("Well, it's complicated…") rewritten to a specific request: "Give me one example of where it got complicated." The lesson: nearly every non-answer traces back to a closed, vague, or leading question.

B. Reading Described Shots

6. Name the fields. The seven fields: Frame, Move, Light, Sound, Cut, Effect, Lesson. In FIGURE 19.3: Frame = MCU, subject on the right third with looking room; Move = locked off at eye height; Light = soft key on the looking side, ~1.5-stop shadow, faint rim; Sound = lav + boom, room tone, no music; Cut = the spine shot, holds on strong lines and L-cuts to B-roll; Effect = spoken-near, not spoken-at; Lesson = the interview look is cheap decisions stacked.

7. Diagnose the frame (model). With the interviewer a full meter to the side: THE FRAME now shows the subject in near-profile, the near eye foreshortened, more cheek than eye visible to camera. THE EFFECT flips from "candid near-eye-contact" to "detached, evasive, watched from across the room." The fix is the whole point of §19.1 — move in beside the lens so the off-axis angle stays small. The subject didn't change; the eyeline geometry did.

8. The pause, dissected. The editor keeps the post-pause line because THE SOUND tells us it arrived after a beat of near-silence — it's the unrehearsed, truer sentence — and THE CUT notes it's a full, self-contained statement that can stand alone. If the interviewer had jumped in with the next question, that silence never opens, the subject never adds the real line, and the film is left with only the composed (weaker) answer. The pause manufactured the keeper.

9. Write your own (model guidance). A strong answer fills all seven fields precisely and then adds the interviewer's contribution. Example closing line: "The interviewer sat close to the lens (tight, warm eyeline) and, crucially, left three seconds of silence after the first answer — which is where the subject's voice cracked and the real moment happened." Grading yourself: did you attribute the moment to a craft choice (eyeline distance, a question, or a silence), not just to luck or the subject being "good on camera"?

C. Building the setup

10. The eyeline dial (model self-critique). Expected finding: clip (a), right beside the lens, feels like near-eye-contact and intimacy; (b), an arm's length away, feels like a normal conversation observed from the side; (c), two meters away, feels detached and can read as evasive (too much profile). Best practice for most documentary work is (a) or close to it. If all three looked the same, your subject probably wasn't actually turning to face you — check that their eyes, not just their head, tracked to your position.

12. Five Ways (trade-offs). Eye-level (a) is the neutral, respectful default. Low angle (b) lends unearned authority/dominance and often an unflattering up-the-nose view — use only if the story wants that subtext. High angle (c) diminishes the subject, makes them look small or vulnerable — rarely right for a testimonial. Key on the looking side (d) is flattering and motivated; key on the opposite side (e) throws shadow onto the looking side and reads "sinister" — a legitimate mood tool, wrong as a default. The trustworthy pick is almost always (a) + (d).

13. Window-key interview (model). Window on the looking side: soft, flattering, motivated key; face well-exposed, gentle modeling. Window behind: face in silhouette (or a blown-out window if you expose for the face). The lesson is §19.2's common mistake made visible — the brightest source belongs in front of the subject. Keep both clips; the contrast is the most persuasive lighting lesson in the chapter.

14. Two-mic safety (model). A complete answer confirms: two independent recordings of the same words on two channels; both monitored on headphones; levels peaking ~-12 dBFS with headroom; 30 s of room tone captured; and — the test — the "failed lav" sentence is fully covered by the safety track. If your safety was just the camera's on-board mic catching a roomy version, that's still a real backup: it proves the words exist somewhere clean-enough to use or to re-time against. The point isn't broadcast quality on the backup; it's that no single failure loses the take.

D. Asking and listening

15. Open it up (model rewrites). (a) "Do you like your job?" → "What do you love about the work, and what wears on you?" (b) "Was it hard?" → "Walk me through the hardest part." (c) "Are you proud of it?" → "Tell me about a moment you were proud." (d) "What does success mean to you?" → "Tell me about a day that felt like success." (e) "You started in college, right?" (leading) → "How did it begin?" Each rewrite is open, specific, and neutral, and each invites a narrated event rather than a word.

16. The silence drill (guidance). The point is behavioral, not analytical: most people find three seconds of silence genuinely uncomfortable and rush to fill it. Success looks like at least one answer where the subject, into your silence, added something truer or more specific than their first response. If they didn't, you likely broke the silence early or signaled "we're done" with your body language — hold longer and keep your eyes engaged.

17. Full-sentence coaching (model). Before: "Six years." (a fragment — useless once the question is cut). After briefing: "I've lived here for six years." (a self-contained soundbite). The three follow-up answers, with the brief in place, should all be liftable sentences. The takeaway you should write down: the full-sentence rule is the single cheapest thing you can do to make raw interview footage editable — thirty seconds of instruction that saves hours in the edit.

18. Fix the interview (model). Sorted by section: Framing/eyeline (§19.1): interviewer three meters away → move beside the lens; the wide eyeline is why the subject looks evasive. Light (§19.2): window in front of the subject behind camera? No — window behind them → silhouette; put the key in front on the looking side, background darker. Audio (§19.3): one lav, unmonitored, peaking near 0 → add a safety mic, put on headphones, pull levels down to ~-12 dBFS with headroom. Question design (§19.4): "you were nervous, weren't you?" is leading → "how did you feel?"; reading from a phone means they're not listening → look up and listen. Listening (§19.5): "right, right" over answers → nod silently; and they should be leaving silences, not filling them. Every problem is a decision, and every one is free to fix.

E. One vs. two cameras

19. Same answer two ways (model). Single-camera: internal cuts (removing an "um," tightening) will jump on the locked medium → you must hide them with B-roll, a 4K punch-in, or a reframe. Two-camera: cut from A (medium) to B (close-up) on the trim and the edit is invisible — provided both cameras are on the same side of the line and ≥30° apart. The exercise should make the trade concrete: two cameras move the effort onto set; one camera defers it to B-roll.

21. Audio plan (model answers). (a) Quiet office: lav (CH1) primary + shotgun (CH2) safety, ~-12 dBFS, monitor both; easy. (b) Busy road: get closer with both mics, use the shotgun's rejection pointed away from the road, consider a quieter time/spot (fix in pre), and definitely record extra room tone of the specific ambience; the lav's proximity is your friend. (c) Gestures + scarf: the scarf will rustle the lav — place it carefully or expose it above the scarf, and lean on the shotgun as the likely primary; brief the subject gently about the scarf. In all three: two mics, two channels, headphones.

22. Match two cameras (model). Match: white balance, exposure, frame rate, shutter, picture profile, and eye height between A and B; and keep both on the same side of the line, ≥30° apart with a size difference. Do two things for the editor: (1) sync them (a clap or timecode) so the angles line up frame-accurate; (2) keep the interviewer beside one camera so the eyeline stays consistent. Skip matching WB/exposure → the A↔B cut jumps in color/brightness (a painful Chapter 31 fix). Skip sync → the editor can't cut between angles cleanly at all.

F. Recreate It

24. Direct-to-lens interview (model). Compared to off-axis (Exercise 10), the direct gaze feels like address/implication — the subject talks to you, the viewer, not to someone off-screen. You'd choose it when you want the audience implicated and spoken-to directly (a personal appeal, a confession, a Morris-style reckoning; see Case Study 1). What it costs: intimacy-with-an-interviewer is replaced by intimacy-with-the-camera, so the subject needs somewhere real to look (a rigged face beside the lens) or they'll look hollow — and misused, direct-to-lens tips into "spokesperson ad." It's a deliberate exception to the default, not a new default.

G. Interleaved

25. Interleave with Ch.7 (model). A two-camera interview is coverage of a single subject, so the same rules apply: keep both cameras on the same side of the subject's eyeline (the 180-degree line) so the gaze direction stays consistent, and offset them by at least 30° (plus a size step) so the A↔B cut reads as a shot change, not a jump. A "jump cut" between two cameras that are less than 30° apart and the same size would look like the subject twitched — the exact error the 30-degree rule (Chapter 9) exists to prevent.

26. Interleave with Ch.10 (model). From Chapter 10's relaxed-performance toolkit, three moves for the first two minutes: (1) chat before you brief — get them talking about something easy and unrelated while you finish setup so the camera stops feeling like an event; (2) run deliberately throwaway warm-up questions (name, what they do) so they hear their own voice and settle; (3) manage your own energy and eye contact — a calm, warm, present interviewer beside the lens makes a nervous subject calm. Never open with the hardest question.

27. Interleave with Ch.13 (model). Seat the subject a few feet from the north window with the window on their looking side (soft, even, cool key — Chapter 13). Camera on the opposite side at eye height, subject on a third facing the window/interviewer. You sit right beside the lens. The lamp goes behind the subject, out of focus, as warm background separation (a practical), not on their face (it'd mix color temperatures on the key). Bounce a white card into the camera-near shadow for fill. One free soft key, one accent — a complete look.

H. Synthesis

29. Teach it back (model, ~200 words). "When you edit an interview, you cut your own questions out — the audience only ever hears the person answering. That one fact changes how every answer has to sound. If you ask 'How long have you done this?' and they say 'Twenty years,' you've got nothing usable: dropped into the film, 'Twenty years' is answering a question nobody heard, so it sounds like a fragment floating in space. You need them to fold the question into the answer: 'I've been doing this for twenty years.' That can stand on its own — you can drop it straight onto the timeline and it makes sense with no question attached. So on set, you do one simple thing: before you roll, you tell the subject, 'You won't hear my questions in the final piece, so answer in complete sentences — instead of yes, say why in a full sentence.' Most people get it instantly, and it turns a pile of fragments into a pile of usable soundbites. The interview that cuts easily isn't the one with the most eloquent subject — it's the one where the interviewer set the answers up to stand alone." (Uses the disappearing question, a fragment-vs-sentence example, and the on-set brief.)

Guidance for the remaining items (16, 20, 23, 28, 30): these are personal/production tasks graded on doing, not on a single right answer. Strong work shows: (20) a genuine 20% tighten with every internal cut hidden and a note on which method was easier; (23) a recreated three-point look critiqued field-by-field against FIGURE 19.5, honestly naming the biggest gap as placement, not gear; (28) a one-page plan that solves light and sound via when and where (a preview of the Production Checkpoint); (30) an honest 1–5 self-audit on the four pillars with one specific, actionable change named before the real Project 2 interview.

Chapter 20 — Answers to Selected Exercises

(Model answers and critiques for the starred and odd-numbered items; the even items marked "answer provided" in the exercise text are included where they teach something specific.)

1 (Count the cutaways). No fixed number — a well-covered three-minute interview segment might leave the talking head 15–40 times. The teaching point is when the cutaways happen: they cluster at (a) the joins between answers (hiding interview trims) and (b) the moments the subject describes something concrete (showing it). Reward a student who notices cutaways land on edits and on nouns.

2 (Find the L-cut). Any moment where the voice continues while the picture shows something else is the target. The strong observation is that the B-roll proved the sentence it played under (voice: "we do it all by hand" → picture: hands working). That double duty — cover the cut, prove the claim — is the whole point of §20.4.

3 (The sequence hunt). A complete sequence shows one action in at least three sizes, usually wide→medium→detail, cut as one. Reward correctly naming the three sizes in order of appearance; note that pros often open on the detail or the medium for a hook and reveal the wide later — a valid variation, as long as all three sizes exist.

4 (Sound-off test). If the picture track carries the story muted, the B-roll is doing its job (and serving sound-off viewers); if it collapses, the piece leaned too hard on the interview and under-shot its cover. The finding either way is diagnostic — this is exactly the test to run on the student's own Project 2.

5 (Reverse the coverage plan). A good reconstruction lists each visible B-roll shot, labels it (cutaway/insert/sequence), and — the hard part — infers the safety-net shots that must exist but aren't foregrounded (an establishing wide, room tone, extra angles). The lesson: what you see in the cut is a fraction of what was shot; the ratio is real (§20.6).

6 (Name the tool). (a) cutaway (a shot away to the empty workshop); (b) insert (a detail within the action — the hand pressing the stamp); (c) a sequence (three sizes of one action cut together). If a student calls (b) a cutaway, gently correct: it points closer, not elsewhere.

7 (Diagnose the field — the steam). THE LIGHT is motivated because the café's window has been the room's key source since Chapters 11 and 13; steam rising in front of it is rim-lit (backlit) and glows, which is why the shot is beautiful and consistent with every other café shot. If the steam were lit by, say, a hard frontal source the scene doesn't have, it would look pasted-in — the light wouldn't match the room, and the cutaway would jar against the coverage it's meant to cover. The lesson: even a B-roll insert obeys the scene's established light.

8 (Rewrite the cut — wide only). With only the wide, the "Cut to next" column collapses: there is no medium or detail to cut to, so the editor cannot punch in on the pour, cannot compress the ninety-second process, cannot cover an interview trim with a detail, and is stuck showing the whole action in real time from one distance. The sequence stops being a sequence and becomes one flat shot. This is the concrete cost of not covering an action in sizes.

9 (Write your own Described Shot). Grade on all seven fields being present and specific, and especially on THE CUT describing how the shot lays over the interview (e.g., "cuts in over the line 'I've done this thirty years,' the interview audio continuing beneath — an L-cut — hiding a splice while proving the claim"). A strong answer's LESSON names a transferable principle, not just a description. Model shape: an ECU of the subject's worn hands, locked off, window-lit, wild sound of the work, cut over the "thirty years" line, effect = the wear proves the years, lesson = objects carry time.

10 (One insert, braced). Critique the student's own clip: is it actually steady enough to hold three seconds without the wobble pulling the eye? Common failures: handheld micro-shake, focus hunting, too-busy background. The fix is bracing (a surface, a mini tripod), locking focus, and simplifying the background. A rock-steady insert is the single most useful B-roll shot to be able to produce on demand.

11 (The five-shot sequence). Self-critique rubric: (1) are all three sizes present (wide, medium, at least two details)? (2) does the same action overlap across the sizes? (3) does at least one cut land on matching action so the motion carries? (4) does it read as one fluid action? The most common failure is not enough overlap — the student shot different moments at each size, so the pieces won't line up. Fix: repeat the whole action on every pass.

13 (Grab the wild sound). The point students should articulate: the picture suddenly feels present and real — a shop that sounds like a shop, coffee that hisses — instead of a silent, slightly dead slideshow. Wild sound is cheap on set and transformative in the mix; it's Theme 2 (sound is half the picture) applied to B-roll, and it's the cousin of the room tone from Chapters 14–15.

15 (The café, your version). Reward whether the student has enough to cover a dialogue scene: an establishing wide, at least two inserts (hands, a detail), a couple of cutaways (a sign, the room, steam or an equivalent), and matched light from a real window. The self-test: could they lay these over an order exchange to hide two trims? If yes, they've grasped the anchor lesson.

16 (Hide the jump). The covered version reads as continuous because the audio runs unbroken while only the picture changes — and the eye accepts a change of picture under a continuous voice (§20.1's "why it works"). Without the cover, both channels lurch at the splice and the jump is naked. Students should feel, viscerally, that B-roll is what makes an interview edit invisible.

17 (Build the sequence two ways). Wide→medium→detail is the "establish, then narrow" default — it orients the viewer before intimacy. Opening on the detail is a hook move: it creates mystery ("what is this?") and reveals context later. Neither is wrong; the trade-off is orientation vs. intrigue. Reward a student who can say why each order serves a different intent.

18 (Lay B-roll over a real answer). Grade on whether the picture proves the words and the audio runs unbroken (an L-cut), and whether a viewer felt picture and words as "one thing." The commonest error is redundancy or mismatch — B-roll that doesn't relate to the sentence (the "beautiful shot of nothing," §20.2) or that just re-states it. Strong answers pick B-roll that shows what the words can't say alone.

19 (Fix the plan). Three problems and fixes: (1) "a couple of pretty shots" is not enough cover — the interview will be full of visible jump cuts; fix: build a coverage plan from the beats and shoot 10× what you'll use. (2) "pretty shots of croissants" risks the beautiful-shot-of-nothing; fix: shoot shots matched to what the baker says (her hands, the 4 a.m. start, the sourdough she mentioned). (3) no sequences named; fix: cover at least one full action (shaping a loaf) wide/medium/detail. Bonus: no wild sound or room tone mentioned.

21 (The unmotivated reel). None of it cuts in because none is of the seamstress or her work — leaves, flares, a spinning cup, and a parking-lot drone shot answer no story beat and cover no specific cut; they're decoration (§20.2's Common Mistake). Rewrite into matched shots: a sequence of her sewing (wide/medium/ECU of the needle), inserts of pinned fabric and worn scissors, cutaways of finished garments on a rail and the empty shop at dawn, a slow reveal of her hands. Every shot now shows a beat the interview will claim.

23 (The steady detail — settings). Start: high frame rate (60 fps+ for silky slow motion), braced hard (mini tripod or a surface), phone/camera inches from the pour with focus locked on the stream, and grab the shot's wild sound separately because slow-motion records no usable sync audio. The trade-off of slow motion: it needs more light (a faster capture) and it plays back silent, so you must plan the sound and expose for the higher frame rate.

25 (Three scenarios). (a) Dentist testimonial — must-have sequence: a patient greeted/treated gently (wide/medium/detail of the calm, clean process); cutaways of the office, hands, a smile. (b) Trail-runner profile — must-have sequence: the run itself (wide of the trail, medium tracking, ECU of feet/breath); cutaways of the landscape, lacing up, the summit. (c) Festival recap — must-have sequence: one vendor or performer's action covered in sizes; cutaways of the crowd, food, faces, the setup. Reward naming one clear must-have sequence per scenario.

27 (Five ways to shoot one detail). Trade-offs: (a) locked-off ECU = clean, reliable, but static; (b) slow push-in = adds life and intent, but risks wobble; (c) high angle = shows the layout/context of the detail, less intimate; (d) backlit against a window = beautiful rim/translucency (great for pours, steam, fabric), but tricky exposure; (e) slow motion = luxurious and tactile, but needs light and records no sound. Which "sells the craft" depends on the subject — backlight and slow motion usually win for anything wet or delicate; locked-off ECU wins for reliability. The lesson: one detail has many treatments; shoot several and choose in the edit.

28 (Recreate the L-cut). Grade on matching the cut structure, not the content: audio unbroken across two picture changes and a return to the head, with the picture cutting at natural pauses/nouns in the speech. The skill is timing the picture changes to the sense of the audio (cut to the thing as it's mentioned), which is the exact L-cut rhythm of professional interview editing (previewed for Chapter 28).

29 (Coverage, old and new). The two kinds of coverage work together: Chapter 7 coverage (the master + sizes of the scene/exchange, respecting the line) gives the editor the scene; Chapter 20 B-roll cutaways (details and room shots around it) give the editor cover to trim and compress that scene. A student should notice that the Chapter 20 cutaways must respect the same line as the Chapter 7 coverage, or they'll flip the space when cut in.

30 (Show the beat). Models: "she's meticulous" → an ECU of her redoing a stitch nobody would ever see, or aligning a seam twice. "The work is hard" → hands cracked and taped, a bead of sweat, the weight of a piece lifted. "It's a family business" → two generations' hands in one frame, a faded photo on the wall, a child's drawing by the register. Any concrete, wordless behavior/object beats the adjective (Chapter 17, §17.4).

31 (Continuity in a sequence). The broken version lurches — the hand teleports across the cut because the action didn't match, exactly the jump the 30-degree rule and matching action (Chapter 9) exist to prevent. The fixed version, cut on the movement with the hand in a matching position, flows as one action. The lesson: a sequence is only seamless if the action overlaps and you cut on the motion.

Chapter 21 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises. Grade on specificity and honest reasoning, not on matching these word for word.

1 (Name the mode). Strong answers name the mode and the concrete clue: "expository — a narrator told me what I was seeing"; "vérité — no narration, no interviews, the camera just followed the action"; "participatory — the filmmaker was on camera asking questions"; "essay — a personal, subjective voice organizing images by association." The lesson: modes are readable from a couple of concrete signals (narration? interviews? maker present? argument or observation?).

2 (Count the narration). The point is to notice what carries the story. In many strong docs the subject's own words (interview/soundbite spine) do more work than any narrator, with pictures and sound filling in. If a narrator carries everything, the film is expository; if pictures and real voices carry it, it's closer to observational. There's no "right" ratio — the skill is hearing the balance.

3 (Spot the shaping). A strong answer names a specific choice and its effect: "they cut from the CEO's promise straight to a shuttered factory — the juxtaposition made an argument no line of narration did," or "the music turned a routine meeting ominous." The lesson (previewing §21.5): meaning is manufactured by selection, order, and music — always, in every documentary.

5 (Find the arc). Grade on whether the four beats are actually present and specific. Good: "Hook — a diver already underwater; Question — will the reef survive?; Turn — the scientist admits her own study was wrong; Resolution — a fragile, qualified hope." If the student genuinely can't find a turn, that's a real finding: many weak docs are flat, and naming the absence is the skill.

6 (Read the vérité shot). Two markers of unmediated feel in FIGURE 21.2: (camera) the handheld, walking, near-continuous take — the camera is a body in the crowd, not a tripod outside it; (sound) live synchronized location sound captured by a portable rig, not staged or re-recorded. Both say "this is happening now, and no one arranged it for us."

7 (Diagnose the still). Pulling out from a single face to reveal the whole crowd changes THE MOVE from intimate-and-narrowing to context-and-widening, and THE EFFECT from "know this one person" to "see this person swallowed by a moment larger than them." Same photo, opposite meaning — proof that the direction of a motivated move is itself a storytelling choice.

8 (Critique a mode mismatch). The execution betrays vérité in every way: a locked tripod (vérité is usually handheld/observational and mobile), a three-light setup (vérité uses available light), and a cued "act natural" performance (vérité captures unstaged reality). Fix: go handheld or minimal, use the room's own light, stop directing, get close, keep rolling, and film real activity — not a performance of it.

9 (Write a vérité Described Shot). Grade on precision and on the vérité discipline: available light, unbroken observation, sync sound, no staging. The one-line reflection is the real prize — e.g., "if I'd stopped the vendor to 'set it up,' I'd have lost the moment he laughed at a customer's joke — the whole reason the shot lives." Vérité's value is the unrepeatable real moment.

10 (Shoot a two-question arc). Guidance: the win is realizing that two good questions ("what are you trying to do?" / "what's in the way?") plus a little B-roll already contain a want, an obstacle, and the seed of an arc. If the answers are flat, the questions were probably closed or the student didn't go quiet after the answer to let the real thing surface.

11 (Cut the spine). A good spine is two or three self-contained sentences that tell the story with the picture off. If an arc appears, the hook is usually the most intriguing line and the turn is the most surprising or emotional one. If no arc appears, the interview lacked a stake or a change — a shooting problem to fix next time, not an edit problem.

12 (Cut it two ways). Both edits must use only honest selection and order — no words the subject didn't say, no manufactured causation. "Triumphant" leans on the confident lines and successful B-roll; "uncertain" leans on the hedges and the unresolved beats. The ethics lesson: this is legitimate interpretation right up until you keep a sentence they didn't mean or imply an event that didn't happen. Have students mark where their cut gets closest to that line.

13 (Find the turn). Guidance: the turn is usually the single most surprising, contradictory, or emotional sentence in the footage — often one the subject almost didn't say. Structuring toward it means everything before it sets it up and everything after pays it off. This is the paper-edit instinct (Chapter 30) — finding the pivot the whole film should be built around.

14 (Shoot a cold open). A strong cold open withholds context and creates a question: a real action mid-motion, or a gripping line with no setup ("They told me to burn it. I didn't."). If the viewer doesn't want to know what happens next, the open is probably explaining instead of provoking. Fix: start later, cut the setup, trust curiosity.

15 (The closet test). The bare room sounds hollow, distant, "boxy" — you hear the walls. The closet sounds close, warm, and dead (in the good sense) — the clothes absorbed the reflections. One line: "the closet made my voice sound like it was next to the listener instead of across a room." This is why the closet is the free VO booth.

16 (Write to picture). The "bad" version describes what we see ("he lifts a hammer and strikes the leather"); it disappears because it competes with the picture and tells us nothing new. The "good" version adds what the picture can't ("that hammer belonged to his father"); it deepens the image. The lesson: narration that describes the visible is dead weight; narration that adds context or feeling earns its place.

17 (Cut the third). After deleting a third of the words, the narration almost always reads faster, cleaner, and more confident — and leaves more air for the pictures and real voices to breathe. The lesson: your first narration draft is nearly always overwritten; the delete key is a narration tool.

18 (Three voices). The external narrator feels authoritative and journalistic — the film is about the subject. The filmmaker's own voice feels personal and participatory — the film is a relationship. The interviewee's own words as VO feel most intimate and trustworthy — the film is theirs. Same information, three different relationships to the viewer; choose by what the story wants.

19 (Settings drill: the VO session). Model: space = a clothes closet (kills echo); mic distance = a hand's width, off-axis (warmth without pops); level = peaks −12 to −6 dBFS (headroom, since a clipped voice is unfixable); two insurance items = 10 seconds of room tone (to patch breaths) and 2–3 takes of each line (for choices in the edit). Reward reasoning over "correct" numbers.

21 (Recreate the Ken Burns shot). Guidance: the win is feeling how much a motivated slow move plus one line of voice turns a dead photo into a moment. Common failures: the move is too fast (seasick), unmotivated (ends nowhere), or the scan is too small (mush at the end of the push). Slow, toward a specific point, on a big scan.

22 (Five Ways: one photo, five moves). The slow push, slow pull, and dead-still versions serve a reflective mood; the fast move fights it — speed reads as urgency or energy, which contradicts reflection. The lesson: match the move (including no move) to the emotional register, and never move just because the software makes it easy.

23 (The rights map). Grade on whether the student sorts each item into a realistic category: genuinely public domain (usable), licensable from a stock/archive house (usable for a fee), personal material with permission (usable with a release), or "can't clear / can't afford" (design around it). The meta-lesson: do this before you fall in love with a clip — clearing rights is pre-production, not a post-production surprise (Chapter 38).

24 (Spot the Frankenbite). The subject actually said, "I was angry at first, but honestly, looking back, they made the right call." Cutting to "I was angry… they made the right call" is defensible only if it preserves the meaning — but dropping "looking back" can distort when they felt each thing. The real Frankenbite line is crossed when you splice separated fragments into a sentence they never spoke, or keep a clause that reverses their meaning. Test: does the cut still mean what they meant?

25 (Fix the lying cutaway). The ethical problem: inserting a nod filmed at another time to imply a reaction to a specific answer manufactures a reaction that never happened — fabrication from real footage. Honest alternative uses for the same nod: as a neutral cutaway to cover a jump cut in a general stretch, or over a moment where the interviewer genuinely was listening — never to fake a causal, emotional response to a particular line.

26 (The music test). Over "someone walking to their car": hopeful = warm, rising strings; sinister = low drone, dissonance; melancholy = a slow, sparse piano. All three are "real" music over a real shot, yet each installs a different truth. It crosses from craft into manipulation when the score creates a feeling that wasn't in the moment — scoring a sympathetic person as menacing, or a neutral event as a crime. Amplifying a real feeling is craft; inventing one is manipulation.

27 (The consent conversation). A strong answer goes far past "sign here": explain what the film actually is and where it will be shown (online? forever? to whom?); that they might be shown in an unflattering or emotional moment; that they can't control the final cut but you'll represent them fairly; that they can stop or ask questions anytime; and — for a vulnerable subject — checking they genuinely understand how public "public" is now. Consent is comprehension, not a signature.

28 (The honesty audit). Guidance: the skill is finding your own closest call and being honest about it. Strong students point to a specific choice — "I cut her pause, which made her sound more certain than she was" — and either defend it (the meaning held) or fix it (restore the pause). The maker who can locate their own nearest brush with the deception line is the maker you can trust.

29 (Paper-edit the arc). Model: a one-page beat list where each beat names its soundbite and its B-roll — cold open (a vérité moment), orient (who/where), question (the stake), three development beats (one idea each), turn (the pivot line), resolution (the answer), button (a resonant last image). If a stranger can read it and feel the arc — especially the turn — it will work on screen. If the turn isn't obvious on paper, fix it there.

30 (Build the 3-minute cut). Guidance: don't polish — prove the arc holds. Success markers: the spine is the subject's own words; B-roll covers every picture cut; narration appears only in genuine gaps; the turn is protected with a little air. If it drags, the development beats are probably making more than one point each (compress to one idea per beat).

31 (With and without). Report the actual split from your test viewer. Typical result: the narrated version is followed more easily (clearer, faster orientation); the un-narrated version is trusted more (feels like the subject's own story). The lesson is that "clearer" and "more moving" are often different versions — and choosing between them, per story, is the chapter's central judgment.

32 (Interview + doc). A fact-seeking question ("how long have you done this?") yields a fact; an arc-seeking version ("what did you love about it that made you stay?") yields a want you can build a story on. Same topic, but the second produces the emotional material a documentary runs on. The lesson: in documentary, interview for the arc, not the almanac.

33 (B-roll for beats). Strong answers show the idea, not the words: over "it's lonely work," shoot the empty shop and the single chair — not something literally labeled "lonely." The skill (Chapter 20 applied) is finding the image that carries the feeling of the line, so the B-roll deepens the soundbite instead of redundantly illustrating it.

34 (Light the interview). Success: a single window as motivated key to one side (Chapter 13), a bounce lifting the shadow cheek, and an off-axis eyeline (the subject looking at you, just off the lens — Chapter 19). It should look intentional — modeled face, catchlight in the eyes, subject separated from the background — not flat or backlit. If it looks accidental, the window was probably behind them or the eyeline was straight down the lens.

35 (Pre-pro the shoot). Model: a treatment that names the subject, the likely arc hypothesis, and the spine question; a shot list with interview questions (hunting for want/obstacle/turn) and B-roll sequences (wide/medium/detail matched to beats); and a one-line arc hypothesis you expect reality to improve. The Chapter 16 planning now carries this chapter's arc thinking — that's the interleave.

36 (Teach it back). A strong ~200-word explanation nails the distinction: shaping selects, compresses, and orders real material to reveal what's true (honest — e.g., trimming a rambling answer to its clear core); falsifying manufactures something that didn't happen (dishonest — e.g., a Frankenbite splicing separate answers into a sentence never said). Grade on whether the two examples genuinely illustrate the line, and on clarity for a true beginner.

37 (Recreate a vérité approach). Grade on discipline, not polish: available light, unobtrusive camera, no direction, at least one unbroken 60-second take, genuine closeness. The prize is the reflection — the student should be able to point to one specific unrepeatable moment (a laugh, a mistake, an unguarded glance) that staging would have killed. If everything feels "performed," the camera was probably too obvious or the student couldn't resist directing; the fix is patience and distance-then-closeness, not a second take.

38 (Fix the failed narration). It feels amateur because the narration subtitles the picture — it describes what we can already see, adds nothing, and never lets an image breathe. The rewrite should cut the descriptions entirely and replace them with what the picture can't say (context, feeling, a fact), then fall silent over the moments that speak for themselves ("now he sighs" becomes silence — we can see the sigh). Strong answers mark at least one stretch where the best narration is none.

39 (The four-mode challenge). No single right answer; the value is feeling the trade-off. Typically the observational cut feels more true and intimate but slower to orient; the expository cut feels clearer and faster but more distanced and "told." The insight to reward: the tension between "true" and "clear" is exactly the choice §21.3 and the Production Checkpoint force — and the right answer depends on the subject and the audience, not on a rule. A student who can articulate why their subject leans one way has understood the chapter.

Chapter 22 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises. Yours will differ — these show the reasoning to check against, not the only right answer.

A. Reading the brief and the brand

1. Output vs outcome (model). Examples: a homepage "explainer" video → output is a 90-second explainer; outcome is fewer support tickets and more free-trial sign-ups. A recruiter's "life at our company" film → output is a culture reel; outcome is more qualified applications. A local restaurant's tour → output is a walkthrough; outcome is more first-time reservations. The tell: the output is a noun you deliver; the outcome is a change the client can measure. Most briefs arrive named as outputs; your job is to surface the outcome underneath.

3. Extract a brand kit (model). For a hypothetical "warm, premium coffee roaster": colors — a deep espresso brown (#3B2A1E) and a cream (#F3E9DA), grabbed with an eyedropper from their site; type — a humanist serif for headings (feels crafted, traditional), clean sans for body; tone — "warm, expert, unpretentious"; visual style — soft natural light, shallow focus, earthy palette, calm/locked camera, real hands and steam. That kit alone dictates: window key not hard light, warm grade, serif lower-thirds, and gentle acoustic music. You just pre-decided half the shoot from a website.

5. The one message (model). For the busy-parent audience, the single message is: "An hour here is the easiest hour in your day — we handle the rest." Cut the pool, smoothie bar, and new equipment not because they're worthless but because a message that lists six features is remembered as zero features; the parent needs one reason that speaks to their real obstacle (time and hassle), and childcare + flexibility deliver it. The other five become supporting B-roll or other videos in a series — not this video's message. Protecting the one message against the client's urge to include everything is the core discipline of §22.1.

6. Write a brief back (guidance). Strong version names an outcome ("get 10 more quote requests a month"), one specific audience, one message, and one CTA, and notes the platform. The useful thing to observe is the gap: owners almost always first ask for a broad "about us" video and, when interviewed, reveal a narrow, urgent goal (a specific service isn't selling; a misconception needs correcting). Writing that gap down — "you asked for X; from our conversation the real job is Y" — is the moment you become a producer rather than an order-taker.

B. Reading testimonials and product shots

7. Applause or proof? (answer). (a) "It's a fantastic product, I highly recommend it" persuades almost no one — it's vague applause, indistinguishable from a script, with no evidence a skeptic can hold onto. (b) The Saturday/accounts line persuades strongly: it's specific (a whole Saturday, once a month, now twenty minutes), it dramatizes a real before/after, and it ends on an emotional benefit (time with the kids) the target audience feels. Specifics are what a peer believes; superlatives are what they discount. The fix for (a) is always a more specific, open question (§22.3, FIGURE 22.5).

9. Read the hero shot (answer). FIGURE 22.8 handles reflection by making the big soft source's reflection become the shaping highlight that curves along the metal, and using black cards to place clean dark edges — the product reflects a designed world, not the room. The depth of field is shallow (f/2.8) to isolate the product from the sweep and direct the eye to the sharpest detail — desirable, premium separation (Ch.4). The single slow motivated push makes a static object feel alive and ends on the key detail (Ch.8). It would read cheap if: the surface reflected the ceiling light, the camera, or the operator; the light were hard and flat; the background were cluttered; or the move were an unmotivated spin.

10. Write your own Described Shot (model). A strong answer fills all seven fields precisely (e.g., for a headphones ad: FRAME — tight three-quarter on the ear cup on a graded sweep; MOVE — slow orbit; LIGHT — big soft top source describing the curve, rim for edge; SOUND — a designed bass swell; CUT — to a detail of the hinge, then hands lifting them; EFFECT — desire/precision; LESSON — orbit + soft shaping = premium) and then names the benefit: not "40mm drivers" but "the world goes quiet and it's just you and the music." The benefit line is the test that you understood §22.4.

C. Shoot this — the testimonial

11. Ten questions, zero closed (model set). For a customer of a bike-repair shop: (1) "Walk me through the bike problem you had before you found them." (2) "Tell me about the day you first brought it in." (3) "What did you expect, and what actually happened?" (4) "Take me through the moment you got it back." (5) "What's different about riding now?" (6) "Tell me about a time they went out of their way." (7) "What would you have done if they didn't exist?" (8) "What do you tell friends about them?" (9) "What would you say to someone on the fence about the price?" (10) "Is there anything I didn't ask that you'd want someone to know?" Must-gets marked: (1) and (9) — the problem and the customer-voiced CTA. None answerable "yes"; several are "walk me through / tell me about the time."

13. Use the pause (answer). Expect the first answer to be the "prepared" one — composed, a little general — and the line after two to three seconds of silence to be more specific, more personal, and more usable. Amateurs fill the silence and lose it. Watching back, you'll typically find your keeper soundbite arrived after a pause, not in the initial reply. The discipline is craft you can practice: say nothing, keep your eyes on them, let it get slightly uncomfortable.

15. Testimonial + B-roll (guidance). Success looks like: you can lay something relevant over every internal cut in the interview — the customer in their world, the product in use, the before/after — so the talking head never sits static long enough to feel like a hostage video (Ch.19 §19.6, Ch.20). If you find a cut with no B-roll to cover it, that's a coverage gap you'd fix by shooting more; note it. The point of the exercise is to feel, before you edit, how completely B-roll depends on decisions made on set.

16. Rescue a scripted one (answer). The scripted version reads as an ad: the eyes flick as they recall the line, the cadence is a reading cadence (even stress, no natural emphasis), and there's no unguarded specificity. The properly interviewed version has live eyes (they're thinking, not remembering lines), natural emphasis and pauses, and at least one specific detail no script would have written. The credibility difference is total — and it's why you never script a testimonial's content (§22.3). Steer structure; ask specifics; use the pause.

D. Shoot this — the product and the demo

17. Find the reflection (answer). Bad version: the object mirrors a hard ceiling light, the window frame, or you and the camera — hard hotspots and a cluttered reflection that read as cheap and accidental. Good version: a single large soft source (a window through a sheer curtain, or bounced) becomes a smooth, gradient highlight that describes the object's shape, with a black card killing distracting reflections on the other side. The lesson you should feel: you don't light the product so much as you light what it reflects (§22.4).

19. Five Ways (answer/trade-offs). (a) Soft light = flattering, premium shape, but can look flat without shadow. (b) Hard raking light = dramatic texture (great for leather, metal, fabric) but harsh on smooth or flawed surfaces. (c) Flooded flat light = even but lifeless — the "shapeless blob." (d) Adding negative fill to (a) = the winner for most products: soft shape plus a dark side that gives 3D form. (e) Slow slider push = adds life and production value if motivated, distracting if not. Most desirable is usually (d) — soft key with negative fill — because form comes from the dark side, not the light.

20. Feature to benefit (model). For a vacuum-insulated bottle: "500ml" → benefit: "a full day's water in one fill" → shot: hands filling it once at breakfast, then drinking from it at sunset. "Vacuum-sealed" → "your coffee's still hot at 3pm" → shot: steam rising when the lid comes off mid-afternoon. "Powder-coated steel" → "survives being dropped in a car park" → shot: it bounces off concrete, unbothered, hands pick it up. Each pairs an invisible spec with a visible payoff — show the hole, not the drill.

22. Settings Drill: the tabletop (model). Premium reflective product: tripod (or slider for one slow push), f/2.8 for isolation on the hero and f/8 for the sharp detail, an 85mm-equiv/macro lens to compress and get close, 1/50 s shutter, ISO 100, white balance set to the key and locked, one big soft key + negative-fill black cards + a rim, and careful reflection control. Rugged outdoorsy tool: swap the light plan toward harder light to reveal texture and grit (a smaller/harder source or direct hard light raking the surface), keep a deep aperture so the whole rugged form is sharp, maybe add a hint of environment (dust, wood) — the brand is "tough," so the light should be too. Same product-shooting principles, opposite mood dial.

E. The call to action and the platform

23. Match the CTA to the funnel (answer). (a) Brand story to strangers = awareness; CTA "learn more" / follow — "buy now" fails because a stranger who just met the brand isn't ready to buy (a first-date proposal). (b) Testimonial = consideration; CTA "see how it works" / "read reviews" / "try it." (c) Offer to past visitors = conversion; CTA "buy now — offer ends Friday." The lesson: the ask must match readiness, or the video wastes its persuasion (§22.5).

25. Reframe for vertical (answer). Cropping 16:9 to 9:16 discards the left and right of the frame, so anything important near the edges (a subject on a third, on-screen text, a product at frame edge) is lost. Fixes: recentre the subject for the vertical frame (or reframe/reshoot with vertical in mind), move text and the CTA into the vertical safe area away from top/bottom where platform UI sits, and check the punch-in still holds resolution (4K → 1080p helps). Ideally you planned the vertical on set (§22.5) with vertical-safe headroom, so this is a crop, not a rescue.

27. Cut this: sound-off test (guidance). Watching muted, note every place meaning lived only in the voiceover or sync audio — a spoken statistic, the CTA said aloud, a joke in the audio. The fix is captions plus on-screen text that carries the message and the CTA visually, and visual storytelling (B-roll, on-screen numbers, Ch.34) that doesn't need narration. If the video collapses muted, it will fail on the feeds and pages where most commercial video plays silent (§22.5). Captions are performance and accessibility.

F. Managing the client

28. Frame the review (answer). Something like: "Attached is a rough cut — I'm asking you to approve the structure and the story: is the message right, is the order right, does it land the way the brief intended? Please ignore the placeholder music and the un-graded color; those are the next stage." This directs the client to review the layer you need signed off and pre-empts panic about things you were always going to fix (§22.6).

29. Translate the feedback (model). (a) "Make it pop" → "When you say pop, what do you want the viewer to feel — more energy, more contrast, a stronger opening? Show me the second it goes flat for you." (b) "Make the logo bigger" → "Do you feel the brand isn't clearly present, or that the ending doesn't land? Let's make sure they remember it's you — bigger logo is one way; a clearer end card might be better." (c) "It feels slow" → "Where exactly does it lose you? Is it the open, or a saggy middle?" (d) "I don't love it but can't say why" → "Let's watch it together and pause the moment your attention drops — that'll tell us." Every one converts a symptom into a locatable, fixable problem (§22.6).

31. Hold the scope line (model). "Love that you're thinking about the trade show and the CEO's LinkedIn — those are great uses. Both are beyond our current brief and budget, so let me send a quick note with timing and cost for the extra cuts and shots, and we'll slot them in. For now I'll keep us focused on delivering the agreed hero and vertical so the launch isn't delayed." It says yes to the relationship, names the scope line without a hint of resentment, and routes the money to the proper conversation (Ch.38) — the exact §22.6 move.

G. Interleaved

33. Pre-produce the brand shoot (guidance). A strong answer produces a one-paragraph treatment (the film's feel), a shot list grouping testimonial + product-hero + demo/B-roll, and a simple schedule (pre-light → interview while the subject is fresh → product tabletop → hands/B-roll → room tone/wrap). The thing to notice: the brief changed decisions the treatment alone wouldn't — e.g., "muted product page" forced captions and sound-off design; "gets better with age" forced the old+new product pairing; the brand kit forced warm light and a serif end card. Pre-pro for commercial work is brief-driven (Ch.16 + §22.1).

35. Write the Project 3 brief (model — strong vs weak). Weak: "A 5-minute video about my friend's coffee shop to show how great it is." (No outcome, no audience, no single message, no CTA, no metric — unshootable.) Strong: "Goal: convert first-time visitors browsing the website into people who come in for a weekend brunch (metric: mentions/covers). Audience: locals aged 25–40 who don't yet know the shop exists and think 'it's just another café.' One message: 'This is your new Sunday ritual.' CTA: 'Come in this weekend / see the menu.' Runs on the site (16:9, muted, captioned) + a 20s vertical for social. Brand kit: warm, neighbourly, unpretentious; brown/cream; soft natural light; acoustic music." The strong version half-writes its own shot list — which is the whole point.

37. Plan Project 3's deliverables and approvals (guidance). A complete answer names the platforms and therefore the deliverable family (e.g., a 90s hero + a 30s cut + a 20s vertical + a master, §22.5), and a one-page approval plan with gates (brief → script/storyboard → rough cut → fine cut → delivery), the number of revision rounds included, and the one-line question at each gate ("approve the structure," "approve the polish"). Test: if a busy client read only that page, would they know exactly what they're getting, when, and what "done" means? If yes, you've done the §22.6 job.

38. The audience letter (model). For the coffee-shop Project 3: "I felt like I'd been let in on a neighbourhood secret — warm, unhurried, the kind of place that's been there for years even though I'd never noticed it. I now think of it as 'my' Sunday spot, not just another café. So I checked the opening hours and went in this weekend for brunch." The letter names a feeling (belonging, warmth), a belief shift (from "just another café" to "mine" — the one message landing), and the action (the CTA fulfilled). Writing the ending you want as a concrete change in a real viewer forces the brief to be specific: if you can't picture the viewer's feeling, belief, and action, the brief isn't sharp enough yet. This is the outcome (§22.1) made human.

Chapter 23 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises. Grade on specificity and honest reasoning, not on matching word-for-word.

1 (The scroll clock). The point is to notice what stopped you and how fast. Strong answers are specific: "a face already mid-word saying 'don't do this' — a verbal hook + a promise," or "a hand doing something impossible-looking — a visual pattern interrupt." For the losers: "a slow logo," "a person clearing their throat," "nothing was happening yet." The lesson: the stop/scroll decision is pre-conscious and lands in ~1–2 seconds, which is exactly the budget the social hook has.

2 (Sound off, then on). Strong answers separate the two jobs: muted, the video must be legible (captions + visuals carry it); unmuted, the sound rewards — a voice, music, an effect that adds energy. If the video made no sense muted, that's the finding: it was built sound-first and will lose the muted majority. Sound is still half the picture; it just can't be the only half in short-form.

3 (Dissect three hooks). A full answer names all three channels for each: e.g., "Visual: a cluttered desk being swept clean in fast motion. Verbal: 'Your setup is why your videos look amateur.' Text: 'FIX THIS.' Open loop: what's the fix?" The lesson: strong hooks fire multiple channels at once so they work whether the viewer perceives the image, the words, or the text first — and all point at one loop.

4 (Safe-zone audit). Good answers point to specifics: a video where the caption sat at the very bottom and got buried under the username/progress bar (bad), vs. one where the text sat in the center band, clear of the UI (good). The lesson: the platform's interface is not negotiable; the fix is always to move important content into the center-safe column.

5 (Frame Log habit). No single answer — grade on whether the student actually did the daily dissection and whether their notes name concrete mechanisms (hook type, open loop, predicted curve dip) rather than vibes ("it was cool"). This habit is the eye §23.2/§23.6 depend on.

6 (Guess the curve). Reasoning to reward: cliff early = weak/slow open; sag where a boring stretch or a tangent sits; bump at the end = a satisfying payoff or clean loop that drives rewatches. The fix should target the worst feature — usually the hook if there's an early cliff. The skill is diagnosing shape from content, which is exactly what you do to your own analytics.

7 (Name the fields). THE CUT teaches that the hook is chosen and positioned in the edit, not filmed first: "This IS the first shot, and it was placed first in the edit, not shot first." The transferable point: your best two seconds usually happen mid-footage; front-loading them is an editing decision, the highest-leverage one in short-form.

8 (Diagnose the mismatch). Failures: (1) slow locked-off wide = no energy, subject small, reads like an ad; (2) the first words are pure throat-clearing ("Hey everyone, so today…") = an exit ramp; (3) no promise, no loop, no on-screen text = nothing for a muted scroller to grab. Rewrite: open tight, mid-action, first words = the payoff/claim ("This one setting is why your video looks flat"), bold on-screen text stating the promise, a hair of handheld energy. Land the loop in under two seconds.

9 (Write your own hook Described Shot). Grade on precision across all seven fields and especially THE EFFECT (which channels fired and what loop opened) and THE LESSON (the transferable move). A strong reflection: "It hooked me visually before I heard a word — the text and the motion did the whole job, which is why it works on a muted feed."

10 (Read the case study). The best one-sentence capture is something like: "Win the viewer's attention honestly in the first two seconds by front-loading your most compelling promise and keeping it — structure, not spectacle, holds an audience." Reward answers that name the transferable principle, not the specific stunt.

11 (Turn on the grid). Practical check: did they enable the grid + level, place the eyes on the upper-third line, and keep the frame actually vertical? The most common miss is a slightly tilted "upright" frame — the level guide fixes it. Reward a straight, centered, upper-third-eyes vertical.

12 (The safe-zone test). What to check: face and any text sit inside the action-safe center column; nothing important in the bottom ~18% or right ~10%; on the real platform the UI doesn't cover the face. If it does, the fix is to recompose centered and higher, not to move the caption to a corner.

13 (Five Ways). Trade-offs: (a) centered-wide reads clean but can feel distant; (b) tight close-up is the vertical sweet spot — intimate, legible on a small screen; (c) low angle adds energy/power; (d) upper-third with headroom is the safe default for a talking-head. (e) is the trap — the face lands in the platform's username/caption zone and gets covered, proving why you protect the center and give the edges away.

14 (Compose for both crops). Success = the single clip, framed centered and loose, yields a clean 16:9 and a clean 9:16 with the subject well-placed in both. The lesson felt in the hand: "compose for the crop" is real — a centrally-composed subject survives the vertical window, while a side-third composition would have been destroyed by it.

15 (Shoot three hooks — the On Set). Self-critique guidance: the winner is usually the one whose promise is clearest and lands fastest, muted. Common finding: the "in medias res" version often beats the "bold claim" version because motion stops a thumb faster than a static talking head — but it depends on the idea. The real prize is realizing you just A/B-tested a hook, which is how serious creators work.

16 (Move the hook to the front). The reordered version almost always wins: leading with the reveal/result/punchline opens a loop the original buried under setup. The principle: front-loading is the single highest-leverage edit in short-form. If the reordered version feels "spoiled," that's usually fine — the loop is "how did that happen?", which the body answers.

17 (Open a loop). Model top pick reasoning: the best hook opens the tightest, most specific curiosity gap for the target viewer. "The lighting mistake that makes phone video look cheap" beats "some lighting tips" because it names a specific pain and implies a fixable error the viewer fears they're making. Rank by specificity + stakes + how badly the viewer needs the answer.

18 (Kill the throat-clearing). Typical finding: the original handed the viewer 3–6 seconds of exit ramp (logo + greeting + "so today…") before anything interesting. Cutting straight to the payoff reclaims all of it. The lesson: on a feed, every pre-hook second is a gift to the scroll; delete it.

19 (Build a loop). Guidance: the seam shows wherever the last frame's composition, motion, or audio doesn't match the first. Hide it by matching framing and subject position at both ends, cutting on a continuous motion, and letting the music/voice carry across the loop point. A clean loop quietly multiplies watch time (the platform counts the replay).

20 (Caption a clip). Check: large, high-contrast, stroke or plate, center-safe placement, synced 1–6 words at a time. Watched muted end-to-end, the clip should be fully followable. If any essential spoken info isn't captioned, it fails the sound-off test — the whole point.

21 (Fix the auto-captions). What to look for: mangled names, jargon, homophones, and mistimed lines. Reward students who count the errors — realizing there were, say, six errors in fifteen seconds is the visceral lesson in why raw auto-captions are worse than none. Correcting them is non-negotiable (♿).

22 (Cut the dead air). Typical result: a 30-second clip drops several seconds and gains energy once breaths, "um"s, and pre-point pauses are cut on the sentence. The pause worth keeping is one that lands something — a beat before a punchline or a reveal. The lesson: "no wasted moment," not "as fast as possible" — motivate the pace.

23 (Muted-comprehension test). Guidance: if the muted viewer can't say what it was about, the captions or visuals failed. Common fix: an essential spoken step wasn't on screen, or the text was too small/late. Revise until a muted stranger gets it. This test belongs at the end of every short-form edit.

24 (Read the spec box). The six: aspect ratio (9:16 — fills the phone), resolution (1080×1920 — native sharpness), frame rate (30 fps — matches motion norms), length (~15–60s sweet spot — retention favors short), loudness (~ −14 LUFS — platform normalizes anyway), codec/captions (H.264/265 high-bitrate .mp4 + captions — quality survives transcode, muted legibility). Every number will change — verify in Appendix H.

25 (Settings Drill: three destinations). Models: (a) full-screen short = 9:16, hook in 2s, burned-in captions, 1080×1920. (b) Instagram main feed = 4:5 — taller than 16:9 to command screen while scrolling, but not so tall it's cropped in the profile grid. (c) horizontal YouTube also going vertical = shoot 16:9 but compose centered/loose so a 9:16 crop survives; deliver both. Reward the reasoning, not memorized numbers.

26 (Repurpose by crop-and-reframe). Success = keyframing the 9:16 crop to follow a moving subject so they stay centered. Auto-reframe tends to fail on hard cuts (it lags the new framing), on two subjects (it can't decide who to follow), and on fast motion (it overshoots). The lesson: the reframe is a real edit; the tool is a first pass to fix.

27 (Repurpose by stacking). Reward choosing the stacked layout precisely because cropping would have lost information — a wide demo, a two-shot, or on-screen text that a tall crop can't hold. Stacking preserves the whole frame and turns the freed vertical space into a home for captions/titles. Right tool for "don't lose the sides."

28 (Reframe audit). Guidance: list every wrong guess — a lagged crop after a cut, the wrong subject followed in a two-shot, an overshoot on fast motion, captions now mistimed to the new frame. Fixing them by hand is the point: auto-reframe is a rough assembly, not a deliverable.

29 (Diagnose from the curve). (a) 100%→30% in two seconds then flat = the hook failed; fix the first two seconds, not the middle. (b) holds then cliffs at 12s = a specific dull stretch or an over-long ending at that timestamp; cut/tighten it or end sooner. (c) gentle decline to 60% + end bump = healthy, with rewatches/a clean loop at the end — keep doing it.

30 (Fix the metric habit). The friend is reading likes (a lagging vanity metric). Point them to average % viewed / completion and the retention curve — the numbers that say whether the video works and where it lost people — plus shares/saves, which best predict reaching new viewers. Likes tell you it was seen; retention tells you what to change.

31 (Post and read). Guidance: reward actually reading a real curve and naming one concrete change for the next video based on its shape (usually "tighten the hook" if there's an early cliff). Real data on your own work is the most valuable exercise in the chapter.

32 (Light meets vertical). Model: seat the subject facing (or 45° to) a window so it's the soft key on the face (Chapter 13); place the camera so the lit face lands in the vertical action-safe center column, eyes on the upper third. The window key can be camera-left or -right as long as the face stays center-safe and well-lit. Sketch should show window → subject → camera with the face centered vertically.

33 (Sound still counts). Two reasons: (1) the unmuted viewer watches longer and is rewarded by clean audio/good voice/fitting music — which lifts retention and completion, the metrics that drive reach; (2) clean audio and room tone let you cut smoothly and add trending/licensed audio without a noisy bed fighting it. Sound is half the picture even when it starts muted.

34 (Story is still the boss). Model: keep the single core idea intact; cut the intro, the second and third supporting points, and the sign-off; front-load the most surprising line; caption it; loop it. The protected element is the one idea and its payoff — everything else is negotiable. Reward a version where the story survives compression, not one that's just "fast."

35 (Coverage pays off vertically). Model using a real shot: because you covered the scene in multiple sizes (Chapter 7) and shot B-roll (Chapter 20), the vertical re-cut can choose the shot that crops best for each beat (a tight close-up often survives a 9:16 crop where a wide doesn't) and use cutaways to hide reframes. A single unbroken wide take gives the vertical editor no options and no cover for the crop.

36 (Teach it back). Model (~200 words) should hit: the frame is a tall column, not a cropped landscape (compose vertically, single closer subject, safe zones); the opening is two seconds against an actively-scrolling viewer, so you front-load the hook; and captions carry the message because most watch muted — with one concrete example (e.g., a talking-head reframed centered, its best line moved to second one, captioned). The through-line: it's a different craft for a different viewing situation, not a resize.

37 (The Production Checkpoint teaser). Model critique dimensions: (1) does it hook in two seconds, muted? (2) is the subject safe-zone throughout the reframe? (3) are captions large, proofread, center-safe, and does it pass the muted test? (4) does it match the vertical spec? (5) does it end on a loop rather than a dead end? The honest weakest-part answer is the real deliverable — naming it is how the next teaser gets better.

38 (Recreate the cold open). Success = the promise/stakes are fully legible (on screen and, if audio, in the first words) by the two-second mark, with zero setup before it and something already in motion in frame one. Common failure: the student still eases in with half a second of "context" — cut it; the interesting part must be frame one. The reflection to reward: "I felt like I was starting too abruptly" — correct, and the point.

39 (Recreate a repurpose you admire). Grade on whether the student correctly identifies the method (crop-and-reframe / stacked / blur-fill) and its reason: e.g., "they stacked it because the demo was wide and cropping would've lost the product," or "they crop-and-reframed because it was a single centered subject with room around them." Reproducing the method on their own clip proves they can read a repurpose, not just watch one.

40 (Recreate a caption style). What makes captions legible: large size (readable on a phone at arm's length), high contrast, a stroke or plate so text survives any background, few words on screen at once, tight sync, and center-safe placement. A strong answer names the specific property that carried the original ("the thick black stroke let white text sit over anything") and confirms the recreation passes the muted-comprehension test.

Chapter 24 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises. Grade on specificity, correct priorities, and the one-take mindset — not on matching these word for word.

1 (Count the cameras). Strong answers name a number and how they identified the master: "at least five angles; the one that keeps returning as a clean wide of the whole court, never reframing, is the safety/master." The lesson: live shows are built on a locked wide you can always cut back to, plus tighter iso angles that take the risks.

2 (Spot the cut point). Good answers tie each cut to a motivation: "cut to camera 2 when the new speaker began" (following the subject), "cut to the crowd on the goal" (reaction), "cut wide when the play opened up" (context). The lesson: live cuts are motivated exactly like camera moves (Ch.8) — a cut with no reason feels wrong.

3 (Find the crossed line). If they find a flip, they should describe it: "on the cut, the interviewer suddenly faced right instead of left — a camera was across the line." If they can't find one (the usual, good case), the answer is that all cameras sat on one side of the 180° line, so screen direction held across every cut. Either way, the lesson is §24.2's rule.

5 (Anatomy of a live moment). Grade on whether they reconstruct coverage, not just describe the moment: "there was a wide that held the whole stadium (safety), a tight that caught the winner's face (emotion), and a roaming camera in the crowd (reaction) — all rolling at once, which is the only reason it could be edited." The insight: an unrepeatable moment is only editable if multiple cameras ran simultaneously.

6 (Name the roles). Master — a locked safe wide; not allowed to reframe or stop. Tight — the close-up on faces/emotion; not allowed to still be hunting when the key moment lands (pre-frame it). Roamer — reactions/details/movement; not allowed to become a second copy of the tight.

7 (Diagnose the rooftop). Simultaneous capture means the band, the crowd, and the street are the same minutes from different places, so they interweave in the edit as one coherent event. On one camera you'd have a single viewpoint of an unrepeatable performance — no reactions to cut to, no way to hide a reframe, no coverage — and you could never "go back" for the missing angle because the moment happened once.

8 (Rewrite the field). THE LIGHT changes from flat grey daylight to a controllable lit stage (you can now expose for a designed look, match cameras to a house wash). THE SOUND changes from raw open-air capture to a clean board feed off the PA. What stays exactly the same: it's still one take, still multicam, still no second chance — the mindset is unchanged; only the inputs improved.

9 (Write your own — the master). Grade on making "still and boring" read as essential. Model THE EFFECT/THE LESSON: "It never wows anyone and never fails anyone — a locked wide can't be caught mid-reframe, so it's the one shot guaranteed to have the vows when everything else went wrong. Its value is that it can't miss." Reward a THE MOVE that explicitly says locked, never touched.

10 (Two angles, one moment — self-critique). The win is confirming in the editor that the two angles actually cut — matched enough (frame rate, WB, exposure) and on one side of the line. The common miss: the tight was on the wrong face at the beat, or the cameras weren't matched so the cut flashed. If they could cut it and it landed on the emotion, they've done real multicam.

11 (Match your cameras). The mismatched version should visibly flash color and brightness on every cut, pulling the eye to the machinery instead of the moment; the matched version cuts invisibly. The lesson (§24.2): matched frame rate, shutter, WB, and exposure are what let the moment survive the cut — the technical discipline protects the emotional continuity.

13 (Three-camera coverage — guidance). Portfolio-grade. Grade on: roles assigned and held (a true master that never moved), all cameras on one side of the line, matched settings, a clap to sync, and audio from the best available source with a backup. The two-minute cut should cut on beats. This is the closest an exercise gets to a real event; the reflection on what nearly went wrong is where the learning is.

15 (Sync by clap). The clap made syncing a filing job — you line up the spike on each angle's audio and everything snaps into place — instead of archaeology (aligning by eye, frame by frame). It's the same self-slate reflex from Ch.18 (§18.3) applied to multicam, and it's the cheapest insurance in the edit.

16 (Cut the moment three ways). (a) Too long on the wide feels distant and slow — we never get close to the emotion. (b) Frantic cutting shatters the moment; no shot breathes. (c) Cutting on the emotion — wide for context, tight on the reaction, held long enough to feel it — serves the moment. The lesson: pace serves feeling; hold the reaction, cut on the beat.

18 (Rescue the missed cut). Fixing it was only possible because every ISO angle was recorded in full — you swap the "wrong" live cut for the right angle from its ISO. If you'd kept only the live program, the missed beat would be permanent. The lesson (§24.3): recording ISOs under a live switch is the safety net that makes live switching survivable.

19 (Fix the single-camera wedding). The plan loses: (1) the vows (the emotion jumps between two faces — one camera can only be on one); (2) the reactions (you're on the couple, not the crying parent); (3) the kiss if you're reframing at the wrong instant. Minimum fix: a second camera — a locked wide that never stops plus a tight — both rolling before the processional.

20 (Fix the streaming plan). Problems/fixes: (1) Venue Wi-Fi saturated → test real upload, go wired, carry a cellular backup. (2) Highest camera bitrate likely exceeds upload → set bitrate to what the connection reliably sustains, with headroom. (3) No local recording → always record the program to disk, so a dropped stream doesn't erase the event. (4) No mention of captions → enable and test live captions.

21 (Fix the audio plan). The problem: a camera-mounted shotgun from the back of the room captures echo and distance, not clean speech — the worst place for the mic (Ch.14 says proximity wins). Fix in priority: (1) get the board feed (venue already has close mics on the speakers); (2) if no board, put your own lav/recorder close to the speaker; (3) always keep a scratch/room mic as backup and sync reference; (4) monitor on headphones the whole time.

22 (Settings drill — match the rig). Frame rate: identical on both (mixed rates can't be fixed). Shutter: 180° for that rate on both (matched motion). White balance: manual, one shared value metered off the room (auto drifts differently per camera → color mismatch). Exposure: matched brightness judged on the waveform (so cuts don't flash). Each must match because you're cutting between angles of the same moment — any mismatch announces the machinery.

23 (The run-of-show — model). Grade on: a real order with timings, the two or three must-gets clearly marked, and a camera assigned to each key moment. Bonus for marking the single moment everything bends around (e.g., "★ the kiss — CAM A wide + CAM B tight, both rolling by [time]"). This is the document that wins the event before it starts (Ch.18 §18.2).

24 (Test your upload). The answer is a real number and a reliable bitrate below it with headroom — e.g., "upload tested at 12 Mb/s, so I'll stream 1080p at ~5–6 Mb/s, leaving room for the connection to dip." Reward anyone who notes upload ≠ download and that venue conditions differ from home.

26 (Diagram your chain — model). Grade on a complete left-to-right chain (cameras → switcher/software → encoder → upload → platform → viewer) with a local backup record branch, and a circled weak link (almost always the upload) with a named backup (wired line + cellular hotspot). The insight is that a chain is only as strong as its weakest link, and you can find it before it breaks.

28 (Recreate the sitcom setup — model). The three simultaneous angles capture a single performance from wide/medium/tight at once — so the genuine one-take energy (and, ideally, a real audience's reaction) survives, and you cut between angles in post. A single re-shot camera would need the performer to repeat the bit for each size, killing the timing and the live reaction. That's exactly why the I Love Lucy method existed (Case Study 1).

29 (Five ways — model). (a) one camera: cheapest, but misses two-person moments — risk: a lost reaction. (b) wide + tight: covers the emotion — risk: mismatch, fix by matching settings. (c) +roamer: adds reactions/energy — risk: crossing the line, fix by staying on one side. (d) live switch: instant deliverable — risk: a missed cut, fix by recording ISOs. (e) stream: reaches remote viewers — risk: the chain drops, fix by local backup + tested upload. For a paying client: usually (b) or (c) cut in post — safety and quality beat immediacy.

31 (The event as a shoot to run — model). Three-camera gala with stream: DP/operator on the tight; a second operator or you on the roamer + audio/board feed; the master is set-and-forget (locked, rolling). Someone owns the switcher + local backup + the stream. Sound is a named owner (board feed + backup, monitored). Two-person collapse: Person 1 = master (set) + tight; Person 2 = roamer + sound/board feed + stream/backup. If you can bring one person, put them on sound (Ch.18 §18.5).

32 (Expose the stage — model). Expose for the lit speaker's face, not the dark audience — let the room fall off; the eye goes to the light anyway. Use the waveform (Ch.5): set the face to a consistent level and match that same level across all three cameras so a cut doesn't flash. Protect the stage highlights from clipping. Matching exposure on the scope is what lets the angles cut without a brightness jump.

33 (The two-camera interview — model). It's a multicam problem: the second (tighter) angle must sit on the same side of the line as the first and be offset by at least 30° (Ch.9's 30-degree rule) so the cut isn't a jump cut; match frame rate/WB/exposure; sync by the common audio (or a clap). In the edit the second angle lets you cut to a tighter reaction, hide an edit in the interview (cover a removed question or a stumble), and add variety — the payoff Ch.30 builds on.

34 (Teach the one-take mindset — model). A strong ~200-word answer hits: there's no retake, so the plan is the performance; the three pillars (run-of-show marks the must-gets so you're framed before the moment; redundancy means two cameras/two audio paths so no single failure loses it; roll early/stop late so you never miss an entrance or a late reaction); and a concrete example (the wedding kiss caught by a locked wide + a tight because both rolled early). The through-line: fix it in pre (Theme 5) taken to its extreme — an event has no post-rescue for a moment you weren't rolling on.

Chapter 25

Chapter 25 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered items. Use them to check an attempt, not to replace it.

1. Spot the specialty shot. A good entry names the technique, the question it answered, and a verdict. Strong: "A drone pull-out ended the travel vlog — it answered 'how small is this island against the ocean?' and earned its place: motivated." Weak-but-honest: "A slow-mo of the host flipping their hair mid-sentence — answered nothing, pure showing off." The skill being trained is the verdict, not the spotting; force yourself to judge motivation, not just admire.

3. Read the interval. Model reasoning: fast-flowing traffic that smears into ribbons → short interval (1–2 s). Clouds that drift smoothly but visibly → medium (3–5 s). A building's shadow crawling while the sky barely moves, or stars wheeling → long (15–40 s). The tell is how much the fastest thing in the frame moves between visible "steps." If motion is silky, the interval was short relative to the subject; if it strobes/jumps, it was long.

5. The dramatic beat. The two-part structure is concealment → revelation ("hide, then show"). The reveal move (fly forward and clear an edge — a rooftop, ridge, or treeline) delivers it: something is hidden by the obstacle, then the obstacle is cleared and the vista appears. That built-in beat is why the reveal is the most satisfying aerial.

6. Diagnose the fields (FIG 25.5). Locked off is required because slow motion multiplies every wobble — at 5× slow, a tiny handheld shake becomes a huge, distracting drift; the subject (falling liquid) supplies all the motion, so the camera must be dead still. Hard backlight does two jobs a slow-mo shot needs: it rim-lights the stream and splash so the fast, thin water reads as bright edges (visibility), and it's bright, feeding the high frame rate so the shot isn't noisy. Handheld + soft frontal light would give you a shaky, flat, and likely dark/noisy 120 fps clip — the three classic slow-mo failures at once.

7. Write your own. A full-credit answer uses all seven fields and ends with the one-sentence storytelling question. Grade yourself on whether THE MOVE names the specific technique (reveal? orbit? hyperlapse? 120 fps?), whether THE LIGHT accounts for the technique's demands (ND for the aerial; brightness for slow-mo), and whether THE CUT places it in a real edit context (an opener? an insert framed by real-time?). If your "question" is "because it looks cool," you analyzed an unmotivated shot — note that as the finding.

8. Your first timelapse (self-critique). Checklist for judging your own: (a) Did it flow or strobe? Strobe = interval too long for the subject. (b) Did it flicker (brightness jumping frame to frame)? = exposure wasn't locked. (c) Did focus hunt? = focus wasn't locked. (d) Was there visible change (sky, crowd, shadow)? If not, the subject was too static — a timelapse needs something that moves. A strong first attempt flows, holds steady exposure, and reveals a change you couldn't see live.

9. The mounted POV (trade-offs). Chest height reads most like "being the person"; low reads fast and aggressive (good for speed, bad for nausea); eye height reads observational. The trade-off is immersion vs. legibility: the more wide/close/fast the mount, the more visceral and the more disorienting. The best take makes a viewer feel the movement while still being able to tell what's happening. If it's just dizzying, it's too wide, too shaky, or too long — shorten it and give it a clear destination.

11. The hyperlapse walk-up. Success criteria: a smooth surge toward a destination with a fixed target held steady in frame. Common failures and fixes: warping/wobble → you drifted off the target (keep one point locked centered) or walked too fast (slow down, smaller steps); jitter → unstable footing (walk heel-to-toe, smooth). The lesson to record: it's the walk-and-talk's "lead your subject" discipline (Ch.8) applied to your own feet — a motivated travelling move with a destination.

13. Five intervals. Model finding: for clouds, 1 s often makes a clip too long and the motion barely-there; 2–5 s usually flows beautifully; 10 s starts to strobe; 20 s strobes hard (clouds jump). The "best" interval is the one where the fastest thing in frame moves a small, smooth step between frames. The exercise teaches that interval is relative to subject speed, not a fixed number.

15. The ground-level reveal. Full credit: the frame starts blocked (an object fills it) and the camera moves to clear it, revealing a subject/vista — reproducing concealment→revelation with your feet instead of a drone. The lesson: the aerial reveal's power is structural, not altitude — you can practice and use the "hide then show" beat with any moving camera, including a phone walked past a doorway or a wall's edge.

17. Punctuation, not paragraph. The 4-second version almost always serves the piece better; 12 seconds outstays its welcome because a timelapse's information ("morning came," "the place filled") is delivered in the first few seconds — after that it's the maker admiring their own shot. The temptation to leave it long is the sunk cost of the shoot (it took 20 cold minutes, so it "deserves" screen time). It doesn't; screen time is earned by the story, not by shooting effort. Shoot generous, cut ruthless.

19. The one-specialty-shot edit. A strong answer justifies the single choice in one story sentence (e.g. "the timelapse opens the piece because the location's transformation IS the story") and then demonstrates the discipline by refusing a second shot it can't justify. The learning is the refusal: budgeting specialty shots in advance is what prevents the demo-reel trap (§25.6). If you found you could justify two, fine — but you had to say why for each, out loud, about the story.

21. The murky slow-mo. Reasons it failed: (a) 240 fps starves for light and candlelight indoors at night is far too dim; (b) the high frame rate forced a very high ISO → noise; (c) the fast shutter high-speed uses compounds the darkness; (d) likely soft because many cameras drop detail at their top rate. Fixes: add a lot of light, OR drop to a gentler rate (60 fps), OR shoot it in real time. Verdict: it should not have been attempted at 240 fps in that light — the honest call is "wrong tool for the conditions."

23. Everything on the action cam. Problems → fixes: ultra-wide distortion on faces/hands → cover with a normal lens/phone at proper distance; no shot-size variety (can't cut, per Ch.7/20) → shoot wides, mediums, close-ups with the main camera; hollow, distant dialogue → the action cam's tiny mic is unusable for speech; record the interview with a proper mic close and off-frame (Ch.14); everything in deep focus → use the main camera's shallower focus to direct the eye (Ch.4). Keep the action cam only for the specific immersive/mounted shots that need it.

25. Slow-mo drill. Models: (a) handshake emphasis → 48–60 fps (≈2–2.5× at 24 fps); barely-slowed, so light is easy — any well-lit room. (b) dog leaping outdoors → 120 fps (5×); the drama wants real slowness, and daylight feeds it — shoot in sun. (c) water balloon burst, max slowness → 240 fps (10×) or higher if the camera allows; this demands bright light (outdoors or heavy added light) or it'll be noisy — the higher the rate, the more light it needs.

27. The pre-flight, from memory. Full list to check against: registration (weight threshold), certification for paid/commercial work, airspace/no-fly (authority map, airports), visual line of sight, altitude ceiling, no flying over uninvolved people/crowds, privacy & consent, insurance, plus conditions/craft (wind, batteries, ND, fly slow). The ones people most often forget: certification specifically for paid work and insurance. The meta-lesson: it's a checklist run every flight, and the rules are local — this is not memorization of facts, it's a habit of verification.

29. The ND thread (Ch.5). Both the cinematic daytime drone shot and the smooth daytime timelapse want a slow shutter in bright light — the drone to keep the 180°-rule motion blur (≈1/50 s at 24 fps) so the move looks natural, the timelapse to give each frame smoothing blur so frames flow instead of strobe. Daylight would badly overexpose at those slow shutters. The ND filter cuts incoming light so you can keep the slow shutter without blowing out — same accessory, same job, two techniques. What they share: both need slow shutters outdoors, and ND is what makes slow shutters possible in daylight.

30. Movement lineage (Ch.8). All three share a motivated moving camera with a destination — the walk-and-talk's essence. What each adds: the action cam adds embodied POV (the camera becomes the subject's eyes); the FPV drone adds speed and three-dimensional freedom (flying through space, not just across the ground); the hyperlapse adds compressed time (traveling through a sped-up world). The most direct descendant is the hyperlapse — it is literally a walk-and-talk executed one frame at a time, sharing the same "travel with/toward a subject" motion and even the same "keep your target framed" discipline; the only thing it changes is the interval between frames.

31. Open a place (Ch.7, 20). Model shot list for a 20-second opening of, say, a harbor: (1) Establisher — a timelapse of the harbor waking (boats leaving, light changing), ~6 s; (2) a wide that grounds the geography at ground level; (3) a medium narrowing to a specific boat/person; (4) a detail insert (hands on a rope, a name on a hull). The establisher answers "where are we / what is this place"; the wide→medium→detail is the Ch.7/20 narrowing that hands off from world to story. Grade on whether the specialty establisher is motivated (does the harbor's change matter?) and whether the coverage narrows deliberately.

33. Justify or cut. The exercise is self-grading: any shot whose best justification is "it looks cool" or "I have the gear" must be cut, and the learning is noticing how many of your beloved shots fail that test. A good response names at least one shot you cut and what you learned (usually: "I wanted it for the shot, not the story"). Keeping a shot is fine — if the one sentence is about the story.

34. Teach restraint back (model). "A drone shot is a tool for answering a question the story is asking — 'where are we, how big is this?' When it answers that, the viewer thinks beautiful place and the shot vanishes into the story; that's success. When it answers nothing, the viewer thinks nice drone shot — and that thought means the story just stopped while they admired your gear. The technique they notice is the one that failed, because its job was to deliver its answer and disappear. That's why Koyaanisqatsi, a film made entirely of timelapse and slow motion, isn't a failure of restraint: every single frame is motivated by one idea, so nothing is showing off — it's all argument. Restraint was never about using few specialty shots. It's about using no unmotivated ones. A film can be all timelapse and disciplined; a two-minute video can have one drone shot that's pure vanity. The measure is always the same: can you say, in one sentence about the story, why the shot is there? If yes, keep it. If your answer is 'because it looks cool,' you have your answer — cut it."

Chapter 26 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises. Students: attempt first, then read.

A. Seeing the cut

1. Count the cuts. Typical average shot lengths: a calm interview or arthouse scene may run 5–10+ seconds per shot; mainstream drama often 3–6 s; trailers, action, and social edits frequently under 2 s, sometimes well under 1 s at a climax. There is no "correct" number — the point is to notice that pace is a measurable choice, and that faster average = more energy.

2. Fast and slow. Model: "The trailer averaged about 1.2 s per shot and felt frantic and exciting; the nature doc averaged about 8 s and felt calm and immersive. The only variable I changed in describing them was cut rate — proof that pace, not content alone, sets energy."

3. Catch a cut you noticed. Common culprits: a jump cut (subject snaps position), a continuity mismatch (a prop moves), a screen-direction flip (crossing the line), or a badly-timed cut that lands in a dead beat. The lesson: visible cuts are almost always mistakes of timing or continuity, not of the transition type. Avoiding them is 90% of looking professional.

5. Dissolve hunt. Model: a documentary dissolves from an old photo to the same location today ("time connects these"); a drama dissolves from a character's face to a memory ("we're entering their mind"). A place a cut was rightly chosen instead: between two angles of the same present-tense conversation — a dissolve there would falsely imply time passing.

B. Reading edits

7. Diagnose the assembly. Drop with least loss: shot 4 (MEDIUM ordering) or shot 5 (pour insert) — the scene survives on wide/reverse/reaction. Cannot drop: shot 8 (the reaction push-in) — it's the payoff the whole scene builds to; without it there's no point. (Also defensible: shot 1, the establishing wide, is near-essential for "where are we?")

8. Re-order for a new meaning. To make it about the café: open on B-roll (steam, machine, window), hold the room longer, and demote the person — put the reaction earlier and briefer, or cut it, and end on a wide of the room. Now the person is a texture in the café's day, not the subject. Same shots, new subject — the Kuleshov lesson at scene scale.

9. Write a Described Sequence. Model checks: the table has # / shot (size+subject) / duration / audio / cut-to-next; the paragraph names where the pace changes (e.g., "shots shorten as the argument heats, then one long hold on the reply lands the emotional beat"). Good answers tie a specific duration choice to a specific feeling.

C. Making cuts

10. The clean top-and-tail. Expected: the trimmed version feels dramatically more "produced." Model sentence: "Killing the 'um, so' at the front and the trailing 'yeah' at the end made a rambling clip feel like a decisive statement — I changed nothing but the in and out points." This is the cheapest quality gain in editing.

11. Assemble a sequence. Success = the four shots play end to end and the action reads (a stranger follows what's happening). Order usually: wide (establish) → medium (the action) → close-up/detail (the payoff). If it doesn't read, the order or a missing size is the cause — note which.

12. Cut it 20% shorter. Almost always: the tighter version has more energy and loses no information — only dead air. The takeaway students should write: "My instinct held shots too long; tightening is not losing content, it's removing waiting."

13. Hide the jump. All three fixes work; the cleanest is usually (a) cutting to a genuinely different size/angle, because it's invisible and needs no extra element. (b) The cutaway is the most reliable universal fix (it always works if you have B-roll). (c) Cutting on a pause is the fallback when you have neither. The meta-lesson: you needed coverage to have options — which is why Chapters 7 and 20 exist.

15. The hold that matters. Success = the held beat lands harder than in a uniformly-paced version. If holding felt "too long" while editing but "right" on playback with a fresh eye, that's correct — editors chronically under-hold their important beats.

16. Dissolve with a reason. The hard cut reads "the next instant / a continuity error" (same place, jarring); the dissolve reads "time passed" and is correct for morning→evening. Keep the dissolve. If you had no handles, you couldn't have made it — note to always roll before/after. This is the §26.3 handles lesson felt firsthand.

17. The transition diet. Typical result: of a dozen fancy transitions, zero to one survive (only a genuine "time passed" earns one). The stripped version looks more professional, not less. Restraint is the craft.

18. Cut on a change. The cut that lands on the movement feels smoother and more motivated; the eye is following the motion and glides across the join. The cut in the stillness is more noticeable. This previews cutting on action (Ch.28).

D. Fixing edits

19. Fix the dead air. Problem: shots held too long (a pacing failure). One-move fix: the tightening pass — take a little off the front and back of nearly every shot; aim ~20% shorter. Drop-off at 0:20 is the classic symptom of a slow first cut.

21. Fix the hostage video. In-edit options: (1) cut to a cutaway/B-roll over the removed sentence to hide the jump; (2) if none exists, cut on a natural pause and accept a small jump; (3) use the fluff's audio under other picture if usable; (4) worst case, live with it. The shooting habit that prevents it: shoot coverage (a second size/angle) and cutaways (Chapters 7, 20) so every internal cut has somewhere to go. The real fix happened — or didn't — on set.

23. Fix the pace curve. Three moves for a saggy middle: (1) tighten the middle third hardest — cut every shot shorter and delete any redundant beats; (2) reorder to move the second-strongest moment into the sag; (3) introduce contrast — a short, faster passage or a held beat — so the middle stops feeling monotone. If there's a music bed, land a change there to re-energize.

E. Setting up and deciding

24. Build the right timeline. Resolution 4K / UHD 3840×2160, frame rate 24 fps, audio 48 kHz — all matching the footage. Wrong frame rate → the editor adds/drops frames to fit its timebase, and motion judders.

25. Match the pace to the brief. (a) energy-drink ad — fast (energy, hook, retention); (b) meditation promo — slow (calm is the product); (c) wedding highlight — mixed, leaning slow/emotional with faster celebratory passages; (d) sports montage — fast/accelerating (build to a climax). Pace serves the feeling the brief wants.

26. Transition decision drill. (a) two angles of one conversation → cut (continuous time); (b) 1990s → today → dissolve ("time passed"); (c) the very end → fade to black (the curtain); (d) hook → first section of a vlog → cut (keep momentum; a transition would kill the pace right after the hook).

F. Recreate and vary

27. Recreate the Kuleshov effect. Viewers reliably read different emotions into the identical neutral face depending on what precedes it (hunger after a meal, grief after a coffin, desire after a person). Their contradictory answers prove the cut, not the performance, made the meaning. Do not skip because Ch.1 named it — doing it is what converts the idea into a reflex.

29. Five Ways: five paces. Findings students should reach: calm and building both "work" for different intents; the single-held-shot version feels intimate or tense; "too fast/frantic" fails because the eye can't register each shot — the viewer feels assaulted, not excited. The lesson: fast is a tool with a ceiling; past legibility, speed becomes noise. Pace must stay readable.

G. The projects

31. Assemble the Café Scene. Success checks: the establishing wide sets place; the covered order exchange cuts between sizes without a jump and honors the 180° line; room tone on A2 runs unbroken so no cut "drops out"; the reaction is held. It should feel like a scene, not a slideshow. Save the project file — Chapters 28/29/31/32/33 all reopen it. Rough is correct; this is a first cut.

33. Coverage → cut (interleaved). The 180° line is an editing rule because its failure only appears in the cut: cross the line while shooting and each frame's screen direction is fine, but cutting two opposite-side angles together flips which way people face, so they seem to look the same direction instead of at each other. The edit can't fix baked-in screen direction — it can only avoid the bad angle, if an alternative exists. Continuity is a promise to the editor.

35. Sound → cut (interleaved). Continuous room tone under both shots smooths the picture cuts because the ear hears no break, so the eye forgives the join — sound "glues" picture. Prediction for letting the dialogue itself lead across a cut: the join becomes even more invisible and more expressive, because we hear the next shot before we see it (or keep hearing the last one after we've moved on). That's the J-/L-cut — Chapter 28.

Chapter 27 — Answers to Selected Exercises

Model solutions and critiques for the starred/odd/flagged exercises. Written for the aggregated Answers to Selected Exercises. Concise but complete; students should attempt before reading.

A. Seeing the workflow

1. Name the pipeline. Offload → back up → ingest → organize → sync → proxy → selects → string-out. Most-missed step: back up (people collapse offload and backup into one) or sync before selects (people try to select from unsynced footage and end up with picture-only keepers). Order logic: safe first, organized before synced, chosen before strung out.

2. The two-copy reflex. Wrong because the shoot now exists in only one place besides the wiped card — one drive failure and it's gone, and the copy was never verified. Correct move: copy to two separate drives with verification, confirm both, then format. "One copy plus a wiped card" is the classic way to lose a wedding.

3. Findability audit (open-ended). Grading intent: any single clip should surface in under ~10 seconds. If not, the fix is almost always a missing or too-coarse bin (e.g., no dedicated Selects bin, or footage not split by scene/camera). Strong answers name the specific bin to add, not "be more organized."

5. The verified copy. The confirmation matters because a file that merely "appears" may be silently corrupt — right name, right size, unplayable. Verification re-reads what it wrote against the source and catches the bad copy while the card still exists, when re-copying is free. Without it, you discover the corruption weeks later, card long since reused.

7. The offload log (model).

CARD | CONTENTS                | A✓ | B✓ | VERIFIED✓ | SAFE TO WIPE
-----+-------------------------+----+----+-----------+-------------
A01  | Cam A interview+B-roll   | ✓  | ✓  | ✓         | yes
A02  | Cam B wide angle         | ✓  | ✓  | ✓         | yes
R01  | Recorder lav+boom+tone   | ✓  | ✓  | ✓         | yes
P01  | Phone grabs              | ✓  | ✓  | ✓         | yes

The value is that "did I copy that card?" becomes a look, never a guess. A card is only wiped when its row is fully checked.

8. Simulate the disaster. Reproduces "media offline": the project links to files by their path, so renaming/moving the source breaks the link. Fix = relink (point the project at the moved files). Prevention (two sentences): decide the folder structure at ingest and freeze it; never move or rename ingested files/folders once a project references them.

B. Ingest and offload

9. Build the tree (model). See FIGURE 27.2 — numbered top-level folders (01_FOOTAGE … 08_DOCS) that sort in workflow order, sub-folders by camera/source under footage, a dedicated ROOM-TONE folder, a disposable 06_PROXIES, and a DOCS folder for brief/releases/script. Grade on: numbering for sort order, room tone separated, proxies kept apart from masters.

10. Mirror it in a project. Bins should match the disk folders one-to-one, plus two project-only bins (Selects, Timelines) with no disk equivalent — because bins hold pointers to clips already filed on disk, not new files; gathering pointers into a Selects bin copies nothing.

12. Bins for a multicam day (model).

📁 FOOTAGE
├── DAY 1
│   ├── Cam A   ├── Cam B   └── Cam C
└── DAY 2
    ├── Cam A   ├── Cam B   └── Cam C   (→ "day 2 / Cam C / speeches" in two clicks)

Divide by the axis you'll search on (day, then camera). Add a Selects bin per key sequence if the event is long.

13. The label pass. The naming choice that makes sorting work is putting the most significant, shared prefix first (project, then setup/scene, then subject/shot, then take) so related clips group when sorted alphabetically — e.g., BIKE_INT_Q3_T04. The full naming scheme (survives across projects and years) is owned by Chapter 37 §37.2; here the rule is just "consistent and sortable, never camera-default alone."

C. / D. Proxies

4. Heavy or light? (a) Light — 1080p all-intra plays easily. (b) Heavy — 4K/60 HEVC (Long-GOP, high pixel + frame load) usually needs proxies, even from a phone. (c) Heaviest — 6K RAW is enormous data per frame; proxies almost mandatory on a laptop. (d) Light — 720p screen recording is tiny. Principle: resolution × bitrate × codec difficulty vs. your machine.

14. The one-question test. The decision is binary and empirical: scrub your heaviest footage; if it stutters/lags/drops frames → proxies; if it's smooth → edit on originals. Don't proxy on principle; proxy on symptom.

15. Spec a proxy (model). Resolution: half (4K → 1080p) — big win, still judgeable. Codec: all-intra editing codec (ProRes/DNxHR flavor or the editor's proxy format) — every frame independent = smooth scrub. Bitrate: low priority — proxies are for cutting, not looking good. Location: 06_PROXIES/ — disposable, apart from masters. Must-do at export: relink to ORIGINALS.

17. Five Ways: playback under pressure (trade-offs). (a) Half-res proxy — best all-round, costs disk + one render. (b) Quarter-res proxy — smoothest, softest to judge focus. (c) Lower timeline playback resolution — instant, no extra files, only affects preview. (d) Render/cache a section — smooth for that section, must re-cache on change. (e) GPU decode — free if your hardware supports the codec, hardware-dependent. On a deadline: usually (a) or (c) — reliable and fast to set up.

E. Syncing

18. Shoot a sync test (guidance). Success = a 20-sec clip with camera scratch audio on, a separate recorder rolling, and one clear clap near the head. This is the raw material for 19–20. Common failure: camera mic muted (nothing to sync against) — the exercise's whole point is to have a scratch reference.

19. Sync by the clap. Method: find the spike in camera scratch audio and recorder audio; slide one until the spikes stack on the same frame. Self-grade by how many frames off the first try was (0–1 frame = good; if lips look off, you're a frame or two out). Play back a plosive/consonant to confirm.

20. Sync by waveform. Auto-sync should match the same clip in a click. Compared to by-hand: far faster across many clips, but fails when scratch audio is too quiet/different — that's when you fall back to the clap by hand. So: auto-sync by default, manual as the reliable fallback. You'd still sync by hand for a single tricky clip or when there's no usable scratch.

21. The no-clap nightmare (open-ended). Expected finding: without a clap or usable scratch, syncing is slow guesswork — nudge, play, judge lips, repeat, with nothing objective to check against. The on-set habit that saves it: keep the camera mic running (scratch) and slate/clap every take. This is the most persuasive argument for the slate a student will ever run.

22. Diagnose a sync fault. Fault: drift — camera and recorder clocks run at slightly different speeds, so a take synced at the clap creeps out of sync over minutes. Mechanism: tiny sample-rate/clock mismatch accumulates. Fixes: (post) re-sync in shorter chunks / nudge the tail; (next shoot) re-clap every few minutes on long takes, use gear with trustworthy clocks, or run timecode.

F. Selects and string-out

23. Circle the takes. Ruthlessness is the lesson: if >50% of clips are "selects," you haven't selected. A select is a keeper, not a maybe. Strong answers flag a small fraction and can say why each earned it (the answer landed / it's sharp / it has life).

24. Carve the good part. Range-selecting the good ten seconds of a two-minute take is more useful than flagging the whole clip because your string-out then contains only usable material — you're not re-hunting the good part later, and the string-out's length reflects real content, not raw runtime.

25. Build a string-out. Success = selects end to end, rough order, no frame-trimming/music, watchable start to finish. The one-sentence-of-story test diagnoses the gap: if you can't state the story, you need either better selects (thin/weak material) or a better order (the pieces exist but aren't arranged to mean anything) — which is the Chapter 29/30 work.

26. String-out to first cut (critique). A good rough cut: beats reordered into beginning–middle–end, obvious dead weight cut, watchable top to bottom, deliberately not polished. The "what next" answer should name a whole-piece issue (pace sags in the middle; the ending is weak; needs B-roll over the seams) — not a single-cut nitpick. Perfecting one scene now is the trap.

G. Fix the workflow

27. Fix: the exported proxy. Cause: the project was left relinked to proxies and rendered from them — a soft, low-bitrate export. Fix: relink to full-res originals, re-export. Prevents forever: "confirm project is on ORIGINALS" as item #1 on the export checklist (Chapter 36).

28. Fix: the vanished footage. Cause: ingested media is linked, not copied into the project; renaming the source folders broke every path → "media offline." Fix now: relink to the renamed/moved files. Forever: freeze the folder structure at ingest; never move/rename ingested files.

29. Fix: the unfindable good take (≥3 failures). (1) Selects never pulled — no flagged keepers, so the good take is lost in the pile → do a real selects pass. (2) Bins/labels too coarse — takes not identifiable → label by scene/take, bin by camera. (3) On the shoot, no slate/take logging (Ch.18) → circle takes on set so the editor inherits a map. Bonus: didn't immerse (watch everything, Case Study 1) → the editor doesn't know their own footage.

H. Interleaving and synthesis

30. The shoot–post handshake (model, three). (1) Clap/slate every take (Ch.18) → pays off at sync (step 5). (2) Record 30–90 sec room tone (Ch.15) → pays off at string-out/first cut (breathing room, hiding cuts). (3) Shoot full coverage — wide/medium/detail + cutaways (Ch.20) → pays off at selects (options to choose from) and hiding seams later. Also: keep the scratch mic on (Ch.15) → sync reference.

31. Codec to proxy (model, ~120 words). Their phone shoots highly compressed HEVC/H.265 — a Long-GOP codec that saves card space by storing mostly the differences between frames. To show any single frame, the computer must decode a whole group of surrounding frames, which is heavy work, especially when scrubbing back and forth. A "professional" camera may record an easier codec (or a lower data load) that the same laptop decodes comfortably, so its footage feels lighter. The fix for both is a proxy: a small, half-resolution copy in an all-intra codec that stores every frame whole, so any frame is instantly available. You edit on the proxy for smooth playback and relink to the original — the real image — only at export.

32. Teach it back (model, ~200 words). "The edit is won before the first cut" means the quality and ease of your edit are mostly decided by the prep work nobody photographs. Before you cut, you ingest the media — offload it off the fragile cards to two verified copies, because a card is transport, not storage, and one lost card can end a project. Then comes media management: you build folders on disk and mirror them as bins in the project, so every clip has a home and any shot is findable in seconds rather than minutes. You sync your double-system audio to picture so sound and image are one, you make proxies if the footage is too heavy to play smoothly, and you pull selects into a string-out. Only then do you cut. Here's the concrete payoff: on a deadline, an organized editor answers "is there a better take?" in five seconds and stays in creative flow; a disorganized one spends five minutes scrolling and loses momentum. Organization isn't the opposite of creativity — it's what protects it. You cannot cut what you cannot find.

33. Build your permanent checklist (open-ended). Grade on: adapted to the student's real drives/editor/naming; run end-to-end on real footage; revised where reality didn't fit FIGURE 27.6. This artifact is the chapter's real deliverable — a career-long tool.

Chapter 28

Chapter 28 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered items. Use them to check an attempt, not to replace it — a cut you got wrong on your own timeline teaches more than an answer you only read.

1. Hear the lead. You'll almost always catch more than two. Typical finds: a phone rings, a door slams, or a new voice starts while you're still looking at the previous shot (a J-cut leading you into the next scene); or a character's line keeps going while the picture has already cut to the person listening (an L-cut). The learning is visceral: with eyes closed you notice that the soundtrack changes on different frames than the picture — which, once you've heard it, you can't un-hear. If you found nothing, you were watching a rare all-straight-cut passage (often a formal news read); switch to any drama or documentary.

3. Spot the cut on action. In a well-cut action scene, most cuts land during motion — a hand mid-reach, a body mid-turn, a foot mid-step, an object mid-flight. What you'll typically log: "cut while the arm was swinging," "cut as she turned her head," "cut mid-stride." The point of the frame-stepping is to catch that the editor did not wait for stillness — the busiest motion frames are the cut points. If you found cuts on still frames that didn't jump, look again: there was probably a sound or a shape bridging them instead.

5. Map a cross-cut's rhythm. The expected pattern: the eight shot-lengths trend downward as the sequence approaches its climax — say 6s, 5s, 5s, 4s, 3s, 2s, 1.5s, 1s — then often a hold or a release at the collision. Your graph should slope down. The insight to record: the editor is shortening the shots to quicken the alternation, and the quickening is the suspense — you feel your pulse rise with the shrinking numbers. If your numbers were flat, either you measured a non-suspense cross-cut (some hold steady for dread) or the sequence wasn't actually building — a useful finding either way.

6. Read the L-cut. The room-tone/music bed on A2 is the continuity floor: because it runs unbroken beneath the picture cut, the ear hears no seam even though the eye sees the picture change to the hands. Without it, the moment the picture cut to B-roll, only the interview voice would carry the sound — and if that voice paused or the B-roll needed its own quiet, the audio would "hole out" and the cut would suddenly become audible. The bed guarantees there's always continuous sound under the picture, so the picture is free to cut. Short version: the bed makes the L-cut bulletproof against any gap in the voice.

7. Diagnose the Café timeline. The J-cut is the barista's line ("what can I get you?") beginning on track A2 under the customer's picture, before we cut to the barista. The L-cut is the barista's voice tailing over the return to the customer. The element that lets the picture cut freely without the sound jumping is the continuous café room-tone bed (A3) — the unbroken ambience recorded per Ch.15. Rewritten THE EFFECT for straight cuts only: "The exchange reads as a series of correct but separate shots — question, then answer, then reply — reported rather than lived; each cut clicks because picture and sound snap together, and the room's sound lurches at every seam." The contrast makes the point: same shots, but grammar is the difference between reporting an order and being in the room.

9. Re-beat the breakfast montage. A strong answer shows each beat at a different stage, with the change legible in the pictures. Model (a friendship cooling): (1) two friends crammed onto one couch laughing, 6s, warm music, cut on a laugh; (2) same couch, a cushion's gap between them, 5s, music thinner, cut on a glance; (3) one texting while the other talks, 5s, cooler, cut on the phone; (4) one arriving late, coats still on, 4s, sparse, cut on a checked watch; (5) the couch with one person alone, 6s, music resolves sadly, end. Grade yourself: does the distance/behavior change across the five beats tell the story with the sound muted? If every beat looks the same, you accumulated instead of compressed.

10. Build a J-cut and an L-cut. Guidance: the J-cut version should feel like the second clip "pulls you in" — you hear the next voice and lean toward it before the picture confirms it. The L-cut version should feel "gentler on the exit" — the first thought finishes in your ear while your eye has already moved on. Both should feel less mechanical than the straight cut, which snaps. If you can't tell the difference, you probably didn't offset the audio enough — try 12–24 frames (half a second to a second) and exaggerate before you refine.

11. Re-cut a dialogue scene, audio only. This is the chapter's central drill; there's no single "right" cut, but the finished version should (a) change no shot and no order, (b) J-cut into each new speaker so you hear them start before you see them, (c) L-cut out so their line tails over the reply, and (d) run one unbroken ambience bed. Self-test: play it with your eyes closed. If it sounds like a conversation (overlapping, flowing) rather than a transcript (one voice, stop, next voice, stop), you've done it. The revelation most people report: the scene feels twice as professional and they added nothing but audio timing.

12. Cut on action, three times. Expected result: version (c), cutting during the movement, is invisible; (a) and (b), cutting before or after on a still frame, "jump." Why: the moving subject captures and carries the eye across the seam in (c); in (a)/(b) the eye is free to notice the shot change, and any small mismatch in the subject's position reads as a hop. The lesson to write down: cut on the busiest motion, not the calm between motions. If (c) also jumped for you, your two sizes probably crossed the line (screen direction reversed) or the positions were badly mismatched — fix direction first.

13. Hide the jump. Success = two nearly-identical framings that would jump on a still cut become seamless once a movement covers the seam. Model approach: shoot yourself seated, same size both takes; in take two, stand partway. Cut on the rising motion — the eye follows you up and never registers that the shot changed. The proof you're after: the same two shots that jump when cut on stillness don't jump when cut on motion. That's the whole thesis of §28.2 in your own hands — motion is a blindfold over the cut.

15. A six-shot montage. Model critique checklist: (a) Are there 6–8 shots, each a different stage? If two shots are interchangeable, one is dead weight — replace it. (b) Is there one continuous music piece, not several? Multiple songs fragment a montage. (c) Do cuts land on the beat? Off-beat cuts feel sloppy in a montage. (d) Mute it — does the change still read from the pictures alone? If not, your shots are too similar. A strong first montage compresses a real change, rides one bed, cuts on the beat, and survives the mute test.

16. Cross-cut two lines. A good answer makes a viewer believe the two lines are simultaneous and converging, without ever showing both in one shot. Success criteria: (a) each line is instantly recognizable (distinct place/person); (b) the shots shorten as the two lines approach their meeting; (c) a viewer, asked, says "it felt like they were happening at the same time." Common failure: the two lines look too alike and the viewer gets confused about who's who — fix with distinct light, location, or a recurring detail per line.

17. Fix the machine gun. Two fixes: (1) stagger the audio — J-cut into new speakers/scenes and L-cut out of them, so picture and sound stop snapping together; and (2) L-cut the voice over the silent B-roll so the cutaways ride continuing sound instead of interrupting it. The B-roll fix is more urgent: silent B-roll is a broken edit (the sound holes out), while an all-straight-cut scene is merely a flat one. Fix the hole first, then de-mechanize the rhythm. Bonus: add one continuous ambience bed to make both fixes bulletproof.

18. Fix the jump. Three ways (this is the interview editor's daily bread): (1) L-cut to a cutaway across the trimmed spot — cover the jump with B-roll (hands, a detail, a listener) while the voice continues underneath, so the seam happens off-screen; (2) cut on a motion at the trim point — if the speaker gestures or turns, cut on that movement so the motion hides the jump; (3) change the shot size/angle enough that it's no longer a near-match (the 30-degree rule, Ch.9) — e.g., cut from the medium to a close-up. The most common professional fix is (1). If none is available, you've learned why you shoot cutaways.

19. Fix the dragging montage. It drags because eight shots of a person running are the same shot — there's no difference between them for the viewer to read as progress or elapsed time, so it accumulates instead of compresses, and forty seconds of "the same" feels endless. Fix (selection): replace the repeats with shots of different stages — day one gasping, week two steadier, a hill conquered, the final strong stride — plus variety of size and angle. The principle: a montage is built from differences, not repetitions. Six varied shots beat twenty identical ones.

20. Fix the confusing cross-cut. The failure is readability: two similar-looking lines (similar dress, cars, roads) mean the viewer can't tell, at each cut, which line they're watching, so the "meanwhile" collapses into confusion. Three fixes: (1) distinct light/color — make one line warm and one cool, or one bright and one dim; (2) distinct framing/geography — a recurring establishing detail per line so each cut re-orients us; (3) distinct pace or a distinct recurring object tied to each line. The rule: every cut to a line must instantly answer "which line is this and where does it stand?"

21. Fix the empty match. What's wrong even though each match is "clean": the chain of round shapes (ball → sun → orange → wheel → clock) matches on shape but on no idea — there's no "both about ___" under any of them, so it reads as a hollow parlor trick, and a string of them is worse, drawing attention to the gimmick. A disciplined editor keeps at most one graphic match, and only if it carries a real idea — e.g., the wheel → the clock, if the piece is about a journey and time running out ("both about time passing on the road"). Cut the rest. The clever shape-rhyme is never the point; the idea under it is.

22. Which cut? (a) busy market from a quiet street → J-cut (let the market's sound arrive first, walking the ear in). (b) hands over an interview voice → L-cut (voice continues over the cutaway — the reason B-roll exists). (c) six weeks of training → montage (chosen stages under one music bed, compressing time). (d) a rescuer racing while a victim waits → cross-cut (intercut the two lines, shortening shots to build tension). Full credit requires the one-sentence why for each, not just the label.

23. Split the edit, on paper. Model sketch (V = video, A = audio):

  V1 [ SHOT A: at the window ########### ][ SHOT B: at the desk ########## ]
  A1 [ rain sound ....................... (L-cut: rain lingers over the desk) .. ]
  A2 [ clock tick begins here (J-cut) ................. under the window ........... ]
                                 ^ we hear the clock    ^ PICTURE cuts to the desk
                                                             ^ rain finally fades out

The rain L-cuts forward past the picture cut (so it lingers into the desk shot), and the clock J-cuts backward under the window shot (so we hear it before we see the desk). Both split edits operate across the same picture cut, in opposite directions — the outgoing sound lingering and the incoming sound leading at once. Label credit: picture cuts at the shot boundary; the two sounds cross it on different frames.

24. The bed plan. A montage is usually held together by a single continuous music piece — the pictures leap across time and the unbroken music is the spine that makes the leaps feel like one arc. A cross-cut is often held by a continuous sound bridge that spans both locations — either a music bed or, powerfully, one line's audio (an organ, a phone ringing, a voice) running under both lines to bind two places into one "meanwhile." Why they differ: a montage has one place/idea moving through time (music suffices), while a cross-cut has two places at once (the bridge must convincingly cover both). Both rely on continuous audio; the montage's job is time, the cross-cut's is simultaneity.

25. Invisible or exposed? No single right answer; grade the reasoning. Model: a fast vlog wants exposed jump cuts — they convey speed, energy, and a personal, unpolished immediacy the audience reads as authentic; hiding every cut would feel oddly formal. A wedding film wants invisible cuts — J-/L-cuts, cuts on action, a flowing grammar — because the story is emotional and continuous and visible seams would feel jarring and cheap. A corporate testimonial wants mostly invisible (L-cuts over B-roll to hide interview trims, smooth entrances) with perhaps a few tasteful exposed cuts for energy — professional polish with a little life. The principle: match the cut's visibility to the feeling the piece wants and the audience's expectations.

27. Five ways to leave a scene. Model evaluation: (a) straight cut = neutral, "and then"; (b) L-cut = gentle, the scene's sound escorts you out — good for emotional or reflective exits; (c) cut on action = energetic, momentum carries into the next scene; (d) graphic match = a flourish, best when the two scenes share an idea (else showy); (e) hard cut to silence = a jolt, powerful for shock or a tonal break, cheap if overused. The "showing off" ones are (d) and (e) when there's no reason for them. The right choice is dictated by what the next scene needs to feel like — grade on whether your pick serves the story or just demonstrates the technique.

29. Shoot-for-the-edit audit. This is one of the most valuable exercises in the chapter. A thorough audit finds concrete blocks: "I can't J-cut into the kitchen scene because I never recorded kitchen ambience (Ch.15)"; "I can't L-cut over the interview because I have no B-roll of what she's describing (Ch.20)"; "I can't cut on the door open because I only shot the wide, not the matching close-up overlap (Ch.9)." For each, the one-line shooting note writes itself: "record 30s room tone in every location," "shoot 3 cutaways per topic," "shoot every action fully in each size." The lesson: the edit's freedom is bought on the shoot, and this audit tells you exactly what to capture next time.

31. The Kuleshov callback. Model: using a neutral face + an object + a place, build (a) a match cut — the face looks off-screen; cut, on the direction of the look, to a distant object the face seems to "see" — and (b) a montage — face, then a series of quick shots (a clock, an empty chair, a packed bag) under music, then the face again. The viewer "reads" an emotion and a story you never showed: in the match cut, that the person is looking at that specific thing (you manufactured the eyeline connection); in the montage, that time passed and something was lost (you manufactured it from juxtaposition). The point, straight from Ch.1: the meaning was made in the gaps between shots, not in any shot.

32. The full Café pass. Model critique checklist for the anchor: (a) Does the barista's line J-cut in under the customer before you cut to them? (b) Does the barista's voice L-cut over the return to the customer? (c) Did you cut on the action of the coffee being handed over (the cup's motion hiding the size change)? (d) Is there one unbroken ambience bed under the whole exchange? (e) Compared to your Ch.26 straight-cut assembly, does it now feel like a scene in a real room rather than a report of an order? If yes on all five, you've proven the chapter's thesis on the book's own anchor: the scene never changed — your command of it did.

33. Name every seam. The exercise is self-grading, and the ratio is the finding. A polished professional scene will have a bridge at nearly every seam (mostly sound and motion, with the rare match or exposed cut for effect); a beginner's assembly will be mostly "nothing." A healthy target for a flowing scene is that the only bare seams are the ones you chose to leave bare. If you find lots of accidental "nothings," you've just located exactly where your scene clicks — and each one tells you which bridge to add (a lingering voice, a motion to cut on, a bed underneath). The habit, not the single audit, is the skill: run it on every project until you're asking "what carries them across?" without thinking.

34. Teach it back (model). "Think of two shots as two rooms. A beginner treats the doorway between them as a wall — the first room just stops and the second starts, so the viewer feels a bump every time, like walking into a closed door. A pro treats that doorway as a hinge: something swings across it and carries you through without noticing the threshold. Sometimes that something is sound — you hear the next room before you see it (a J-cut), or the first room's voice keeps talking as you step into the second (an L-cut), which is how a shot of someone's hands can play under their story. Sometimes it's motion — you cut in the middle of a reach or a turn, and because your eye is chasing the movement, it leaps the doorway with the hand and never sees the edit (cutting on action). Example: film yourself standing up in a wide shot and again in a close-up; cut while you're rising and it's one smooth motion, but cut once you've stopped and you'll visibly 'jump.' The lesson: editing isn't laying clips end to end — it's working the hinges so the viewer flows from thought to thought and never feels the wall." (Grade yourself on whether you explained why it flows — a continuous thing carrying the eye/ear across — not just that it does.)

35. The before/after reel. No single right answer; the value is the honest four-part grade and acting on the weakest score. Typical first-timer result: L-cuts and the bed land easily (big wins), J-cuts get forgotten (people fix exits but not entrances), and cutting-on-action is the hardest (it depends on whether the overlap was shot). If cutting-on-action scored weakest, the fix is usually on the shoot, not the timeline (Exercise 29) — which is the real lesson. Keep the pair; re-graded after Chapter 30, the "after" you're proud of today will look loose, and that itch to re-cut it is proof your eye has grown.

Chapter 29 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises. Many exercises are open-ended shooting/cutting work; for those, the "answer" is a rubric of what a strong response demonstrates.

A. Seeing the cut

1. Which of the six? (Expect this pattern.) Across five cuts, most students find the overwhelming majority serve emotion and story — that's the point of Murch's weighting (74% combined). Occasionally a cut is clearly serving rhythm (a snappy comedic beat) or eye-trace (a cut that keeps a moving object under the eye). Almost none will be primarily about the bottom two — screen plane and spatial continuity are usually served silently in the background, not the reason for a cut. Strong answers name a specific priority per cut and justify it.

2. Count the blink. (Model.) Students consistently discover their felt "ready to move on" moment lands within a beat of where the editor actually cut — often on the completion of a spoken thought or the end of a gesture. The lesson to draw out: the cut point is not arbitrary or metronomic; it tracks the viewer's completed thought (Murch's blink). Where a cut felt "late" or "early," it usually violated that instinct.

3. The held shot. A strong answer times the hold (it's often surprisingly long — 5–10+ seconds), names the specific feeling it produced (dread, tenderness, tension), and explains that cutting sooner would have "resolved" the moment before the audience felt its weight. Bonus: the student connects it to the exhale (§29.3) or the emotional cut (§29.2).

4. Rhythm map a scene. (Model.) A good map shows variation, not uniformity — a cluster of long shots (setup), a run of shortening shots (build), and at least one long hold (exhale). The insight to reach: the scene's feeling correlates with the shape of the durations, not the content of any one shot. A flat, even map paired with an exciting scene means the excitement came from elsewhere (performance, music) and the cutting is underusing rhythm.

5. The mute test in the wild. (Model.) The best answers distinguish two cases: (a) the scene survives the mute (music is amplifying an emotion built into picture and performance) — the healthy case; (b) the scene collapses (music is carrying an emotion the pictures don't) — which the student should recognize is fragile filmmaking, common in ads and montages. The takeaway is diagnostic: if muting your own work kills it, fix the edit, not the track.

6. Frame Log, one week. (Guidance.) No single answer. A strong week shows a shift from what happened to why it cut there — earlier entries describe content, later entries name mechanisms (an L-cut, a held reaction, an accelerating build). That shift is the skill developing.

B. Reading sequences

7. Name the fields. (Answer.) THE MOVE (locked off) contributes stillness that hands the feeling to the viewer instead of dictating it. THE SOUND (dialogue stripped, room tone + low bed) makes the silence carry the moment. THE CUT (cut in on the feeling, hold past comfort, out before resolution) is the emotional decision. The camera is locked off because any move would tell the audience how to feel; stillness lets them do the feeling themselves.

8. Rewrite for the wrong priority. (Answer.) THE CUT becomes: "cut away on the read to a clean insert of the phone, then to the hand on the cup for coverage, then a wide to re-establish geography." THE EFFECT becomes: "the scene is continuous and legible — and emotionally inert, because we left the face at the exact instant the feeling arrived." What's lost: the emotion (priority 1) is sacrificed to serve information and continuity (the bottom of the list) — the textbook wrong trade.

9. Diagnose the montage. (Model.) The three engines: (1) accumulation — many short kisses whose sum exceeds any single shot; (2) a single evolving theme (the Morricone score) binding them into one emotional arc; (3) the reaction — cutting to the watching man's face so we feel through him. A five-song playlist would fragment the sequence into five moods with no build; the unity of one rising theme is what turns a list of clips into a single swell of feeling.

10. Write your own emotional cut. (Model.) A strong seven-field Described Shot nails THE CUT and THE EFFECT specifically: it names the exact in/out logic ("cut in as the eyes lift, hold through the breath, out before the smile completes") and the mechanism ("withholding the resolution makes the viewer lean in"). Weak answers describe the content of the shot but not why it cuts on those frames — the whole point of the exercise.

C. Finding the story

11. The one sentence. (Strong vs weak.) Weak: "It's about my grandmother's bakery." (a subject) Strong: "It's about a woman keeping her mother's recipe alive one loaf at a time." (a point, with a feeling and a stake). The test: a strong sentence tells you what to cut — anything that doesn't serve "keeping the recipe alive" is trimmable.

12. Paper edit first. (Guidance.) Strong responses report that the paper edit surfaced a structural decision (what to open on, what to cut) before the timeline made it expensive — e.g., "I realized on paper that my best moment was buried at the end, so I led with it." The value is catching the shape cheaply.

13. Two sentences, two films. (Answer.) The two cuts should differ in what they keep and hold, not just in captions. "A person finds calm" holds the sit-down and the sip; "a person is interrupted" compresses the calm and holds the phone/reaction. Same clips, opposite emphasis — proof that the editor's sentence, not the footage, is the creative act.

14. Story-cut the Café Scene. (Model walkthrough.) Strong work mirrors FIGURE 29.2: compress the walk-in and order (setup), let the sit-down breathe, and hold the reaction (the point). Success criterion: when shown to a viewer, their answer to "what was it about?" matches the student's one sentence. If the viewer says "someone ordered coffee," the setup wasn't compressed enough and the reaction wasn't held long enough.

15. Chronology vs story. (Guidance.) The re-cut (leading with the most charged moment) almost always beats the strict chronology. Chronology flattens because it gives every beat equal weight in the order it happened, burying the peak; a story assigns weight by importance, not time. Strong answers name the specific moment they moved to the front and why.

D. Rhythm and the breathing edit

16. Kill the metronome. (Answer.) The varied re-cut feels "alive" and the even one feels "numbing" or "mechanical." The principle: an unvarying pulse stops registering as a pulse (like a clock you no longer hear); rhythm requires deliberate variation — some shots snapping, some lingering.

17. Five Ways: one moment, five rhythms. (Trade-offs.) (a) even = neutral/dull; (b) accelerating = rising tension/excitement; (c) decelerating = calming/settling or dread; (d) accelerate-then-hold = a climax that lands (usually the winner for a dramatic moment); (e) one long take = intimacy/realism but risks flatness. The "best" depends on the one-sentence intent — the exercise's real lesson is that rhythm is a choice matched to feeling, not a default.

18. Build the exhale. (Answer.) With the hold, the peak "lands" — viewers report they finally felt it. Without it, the sequence "just ends" or feels "rushed." The exhale gives accumulated energy somewhere to discharge; speed builds tension, but only stillness delivers weight.

19. Cut against the beat. (Model.) Rigid on-every-beat cutting feels mechanical and, past a few seconds, deadening — the viewer predicts every cut. Letting most cuts hit beats while a few deliberately hold past or clip early keeps the edit "human" and alive; the syncopation is what makes it feel authored rather than automated. The lesson: use the beat as a reference, not a cage.

20. Find the right frame. (Guidance.) Students discover a single frame where the cut feels "inevitable" — usually on the completion of a motion or a thought (the blink). Cutting a few frames early feels abrupt/anticipated; a few late feels sluggish/missed. The exercise trains the frame-level sensitivity that separates a fine cut from a rough one.

E. The fine cut

21. Enter late, leave early. (Answer.) Almost nothing is missed — students typically find they cannot name a single trimmed frame the viewer would want back. The trims were "fat" (setup run-up and wind-down). The rare exception is an emotional beat that got clipped; that frame should be given back, which is the whole "without gutting" caveat.

22. The 15% pass. (Model.) A strong log shows the time came almost entirely from setup and transitions (shot heads/tails, entrances, exits), while the emotional beat was preserved or lengthened. If a student hit 15% by trimming the payoff, they optimized the wrong thing — flag it.

23. Fix the gutted scene. (Answer.) Diagnosis: the editor trimmed living silence (the held reaction, the pauses that carried feeling) as if it were dead air, chasing "pace." The scene is faster and colder because the emotion lived in the beats that were cut. Fix: restore the reaction's length and the meaningful pauses; find the fat elsewhere (setup, redundant beats). Speed was never the problem; the wrong frames were trimmed.

24. Fix the flat scene. (Answer.) Likely cause: the cut serves the bottom of the Rule of Six (clean continuity) and neglects the top (emotion) — correct but unfelt. Three things to try: (1) hold on the most emotional face longer instead of cutting to inserts; (2) re-order to lead with the strongest feeling; (3) find the scene's one sentence and re-cut toward it. Correctness isn't the goal; feeling is.

25. The two-minute lock. (Model.) Strong "lock notes" name genuine trade-offs, not tasks: e.g., "held the reaction two seconds longer than felt safe (chose emotion over pace)," "cut my favorite B-roll because it stalled the build (killed a darling)," "let one continuity bump stand because fixing it broke the rhythm (sacrificed the bottom of the Six)." Naming trades = editorial maturity.

F. Music and the emotional edit

26. Place the entrance. (Answer.) The entrance on a clear story beat (a sit-down, a first B-roll shot) feels motivated — as if the moment summoned the music. The random-frame entrance feels "bolted on"; the very-start entrance often pre-empts the scene's own build. The principle: an entrance that lands on a moment reads as intentional.

27. Score it, then strip it. (Model.) The valuable answers are the ones where the scene collapsed on mute and the student correctly prescribes an edit fix (hold the beat, tighten the setup, re-order to the peak) rather than a music fix (bigger swell, louder). Recognizing that a dead-on-mute scene is an editing problem, not a scoring problem, is the whole exercise.

28. Dip for the line. (Answer.) The dipped version lands harder. Reason: contrast, not volume, creates emphasis — dropping the bed under the key line (or key silence) makes it "pop" out of the texture. A constant-level bed flattens everything, including the moment you most want heard. (Real levels/mix: Ch.33.)

29. Settings Drill: place a bed. (Model.) (a) Doc profile: one evolving theme; enter on the first story beat; sit well under narration; swell on the emotional turn. (b) Social hook: music from frame one, punchy and rhythmic; cut picture to the beat; no dialogue to protect. (c) Somber testimonial: sparse bed or none until the hard line; drop out entirely under the line so the silence carries it, then swell after. The through-line: the bed serves the piece's job, and it always yields to the voice.

30. Align the swell to the peak. (Model.) Aligned, the sequence feels "inevitable" and moving — picture and music resolve together. Misaligned (swell over a nothing beat), it feels "off," even faintly comic, and the real peak lands unsupported. Students consistently report the alignment mattered more than they expected — proof that music is structural, not decorative.

G. Killing your darlings

31. Name your darling. (Answer.) The honest answers admit the reason is cost/vanity/effort/ego ("it took all day," "it's my prettiest shot"), not emotion/story. Naming the real reason is the first step to being able to cut it. If the reason genuinely is emotion/story, it's a keeper — not every beloved shot is a darling.

32. Kill it and watch. (Model.) The near-universal discovery: the piece is not meaningfully worse, and often clearer and tighter, without the darling. The emotional shock of "I don't miss it" is the lesson. Full credit requires that the student actually saved the shot to an outtakes bin (nothing wasted) and rendered an honest verdict, even if the verdict is "keep it" for a genuine story reason.

33. Kill the expensive shot. (Model.) A strong argument turns on the Rule of Six, not the budget: "This drone shot opens the geography and delivers the 'scale' the story needs — keep (serves story)," vs. "This crane move is gorgeous but the story is about intimacy; it's here because it cost $400 — cut (sunk cost)." Any argument that rests on what it cost is, by definition, the fallacy the exercise targets.

H. Interleaved and synthesis

34. Sound-led emotion. (Answer.) Leading with sound (the L-cut) softens and emotionally connects the two shots — we feel the outgoing moment bleed into the incoming picture. It serves emotion (#1) and rhythm (#3): the seam is felt rather than noticed, carrying feeling across the cut. A hard cut with synced sound would feel more abrupt and informational.

35. Cut to the arc. (Model.) Strong work shows the rhythm matching the arc: the hook is tight and immediate, the middle breathes, and the resolution holds. A mismatch (e.g., a frantic rhythm over a reflective ending) is the diagnosis to catch. Connect back to Ch.17: you can only cut to an arc you've named.

36. Hide the seam. (Answer.) The cutaway or L-cut should do double duty: hide the jump (continuity, bottom of the Six) and serve the emotion (a cutaway whose content matches the feeling of the line). A cutaway that only hides the seam is a missed opportunity; the best patch also advances the feeling — exactly the move modeled in Case Study 2.

37. Teach it back. (Model.) A strong 200-word explanation: states the six in order with weights; gives the arithmetic (51% > the other 49%); uses one concrete example (a held reaction vs. a "correct" cutaway); and lands the thesis — a viewer feels a cut before they analyze it, so serving the feeling beats serving the match. Clarity and the correct ordering are the grading criteria.

38. The self-audit. (Guidance.) No single answer. Strong work identifies the lowest-scoring priority honestly (often emotion or rhythm, rarely continuity) and re-cuts to raise it without lowering others — demonstrating that the six are levers to balance, not a checklist to tick.

39. The Kuleshov cut, weaponized. (Model.) The same neutral face, after (a) a smashed jar, (b) a child laughing, (c) a ringing phone, reads as grief, warmth, dread — proving meaning is made in the cut (Ch.1). Full credit: the student wrote the one-sentence intent first and chose the preceding shot that serves it, letting emotion lead the footage rather than the reverse.

40. Assembly vs fine cut, side by side. (Guidance.) The viewer's win/lose marks should diverge at specific, nameable decisions — the fine cut wins where it led with feeling, held the beat, and cut the darling; the assembly loses where it opened on setup and buried the peak. The point: the difference is decisions, not footage. Keeping the before/after is the student's clearest evidence of growth.

Chapter 30 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered items. These are models, not the only right answers — an interview edit has many valid cuts. Grade your own work on whether the reasoning holds.

A. Seeing and hearing the craft

1. Close your eyes. A working spine sounds like a story, not a list: it hooks you in the first seconds, has a person and a point, moves through a middle, and lands. If, with your eyes shut, you can say what it's about and you wanted to keep listening, the spine holds — the pictures are only making a finished story better. If it drifts, repeats, or confuses with the eyes closed, no B-roll is saving it; the spine is broken. (This is the eyes-closed pass of §30.6, run as a viewer.)

2. Sound off. A piece that shows a story muted has matched B-roll — you see the person, the work, the place, the beats. A piece that collapses into a static head and a few random cutaways fails the sound-off test and fails every deaf/HoH and sound-off viewer. Most polished testimonials pass both; most amateur ones pass one at best. Note which your chosen piece passed — that tells you where its editor spent effort.

3. Count the seams. Typical 60-second interview-driven stretch: 4–8 cutaways. For each, the job is one of: hide a cut (the audio splices under it), prove a claim (it shows what the voice describes), or both (the best ones). Well-cut pieces are mostly "both." If you find a cutaway doing neither — pretty but covering nothing and proving nothing — that's the beautiful-shot-of-nothing (Ch.20).

5. Reverse-engineer the spine. You should end with a short ordered list of paraphrases that reads as a story. The reorder markers you found (places where a later-life reflection opens the piece, or where a fact is delayed to the middle) are the editor's radio edit showing through. The lesson: every finished doc is a reordering; the chronology you feel is a construction.

B. The radio edit

6. Mark a transcript. Typical keeper-to-mush ratio on a real interview page is roughly 1:1 to 1:3 — most of any interview is throat-clearing, tangents, hedges, and non-answers. If you kept nearly everything, you're not marking hard enough; if you kept almost nothing, either the interview was weak (a Ch.19 problem) or you're being too harsh. The keepers are self-contained sentences that mean something read cold.

7. Deal the cards. A strong ordering opens on the most compelling idea (often the thesis or a vivid reflection), not on "so I was born in...". Model justification: "My hook is the line 'you can't rush bees' because it states the person's whole philosophy in six words and makes a viewer curious what kind of person thinks that — the origin can wait, because who they are is more interesting than when they started." Weak version: opening chronologically because that's how it was spoken.

8. The late hook. A late line makes a good hook when the subject, now relaxed, said something more honest, more distilled, or more surprising than their careful early answers — the throwaway that is secretly the thesis. It works as a hook because it is specific and intriguing and it promises the viewer a person worth listening to, and because it can be delivered without any setup (a full, self-contained sentence).

9. Radio-edit your own. A strong spine: 5–7 soundbites, each a distinct beat, in an order that builds (hook → who → stakes → low/turn → button), with no two bites saying the same thing. A weak spine: everything kept, chronological order, two or three bites making the same point, no clear hook or button. Test: read your spine aloud as a list of paraphrases — does it sound like a story or an inventory? If an inventory, cut and reorder.

C. Selects and the string-out

10. Trim on the breath. A good soundbite starts a hair before the first word (so you don't clip the attack) and ends a hair after the last (on the out-breath), with a little clean room tone on each side to cut on. If it stands alone — states its own subject, needs no question — it's a select. If it only makes sense with your question, either re-cut to find a self-contained version or note it for a re-ask.

11. Label by content. Success = a stranger reading your labels (SB_why-i-started, SB_the-hard-winter, SB_worth-it) can guess the arc of your film. If your labels are Take04, Clip12, they can't, and neither will you at 2 a.m. Content labels turn footage into a deck you can deal.

12. Radio string-out. With eyes closed it should hold as a story. If it drags, the usual culprit is order (a beat is in the wrong place) or redundancy (two bites saying the same thing). Fix by reordering or cutting bites — not by re-wording within them (that risks a frankenbite). Re-test until it holds. Only then are you allowed to add picture.

13. Kill three keepers. Cutting to five usually makes the story clearer, not poorer — each removed bite was costing you time and diluting the point. Name what each cost: "the second 'I love it' bite was a delay before the stakes"; "the equipment tangent was interesting to me and irrelevant to the viewer." If losing a bite genuinely broke the story, it wasn't a darling — put it back; but that's rare. The honest shock of not missing them is the lesson.

D. Building the spine

14. Cut the question out. If the footage has a self-contained version elsewhere, use it. If it doesn't ("...Twenty years." with no full-sentence take), you can't honestly manufacture one in the edit — the fix was on set: you'd have re-asked, "Say that as a full sentence for me — 'I've been doing this twenty years.'" Note it as a Ch.19 lesson for next time; do not frankenbite a subject into a full sentence they never spoke.

15. Snip the "um." Yes — the head jumps on the snip frame, because you've joined two non-adjacent moments of one continuous shot. Note the exact frame. This is normal and expected; it's a seam to cover in Section E. (If you shot two-camera, you'd hide it by cutting to the other angle instead.)

16. The honest tighten. A good 40→12s tighten keeps the one clear thought and trims the wind-up, the tangent, and the "does that make sense?" The moment to watch for: any join where the trimmed version implies a cause, target, or meaning the subject didn't intend. Model reflection: "I nearly joined 'I was exhausted' to '...but I finished' across two unrelated stories, which would have implied they finished that thing — I undid it. The line I won't cross is changing what they meant, only how long they took to say it."

17. Diagnose the frankenbite. Joining "I was furious" (about a shipment) to "...but the client was right" (about a design note) fabricates a causal, conciliatory statement the subject never made — it implies they were furious and then conceded on the same issue, which is false. It's a frankenbite. Honest fix: keep the two statements separate and in their true contexts, or don't use them together at all. The test fails: the subject, watching, would say "I never meant that."

E. Layering B-roll

18. Cover one seam. The seam vanishes because the audio keeps playing (an L-cut) while only the picture changes — the eye accepts a change of picture under a continuous voice, so the interview splice underneath is hidden. Your one sentence should name the L-cut: "the interview audio runs unbroken while the picture cuts to B-roll."

19. Match B-roll to the line. Matched B-roll (hands over "by hand") both hides the cut and proves the claim. The deliberate mismatch (say, the truck over "by hand") still hides the cut but proves nothing — and worse, it distracts, because the viewer's eye tries and fails to connect picture to words. The cost of mismatch: you spend a cutaway and get only half its value, and you slightly confuse the viewer.

20. Time the handoff. Holding the face through the line and cutting away on the tail lets the viewer see the person say the important thing, then rewards them with the proof — the emotion lands, then the picture illustrates. Cutting away early, over the line itself, robs the line of the face at the exact moment the expression is the point; the line goes to a cutaway and loses its weight. The late handoff serves the moment. (This is FIGURE 30.5's lesson.)

21. Cover the whole spine. Success = the interview audio never stops, every seam has B-roll over it, and no jump is visible. The seam you missed is almost always an internal one (an "um" snip inside a bite, not a bite-to-bite join) — those are easy to forget because you're watching for the big joins. Find it by scrubbing the V1 (face) track alone and watching for any un-covered jump.

22. The Café connection. Model answer: "In the Café Scene I laid the steam (rim-lit by the window) over the moment the customer says 'a flat white, please,' as an L-cut, to hide the join between two takes of the order exchange. That's one dialogue cut covered by one cutaway. In an interview, I do the same move — voice continues, picture roams to matched B-roll — but over every seam in a whole spine, and my B-roll now also proves each claim, not just covers the join. Same L-cut, scaled from one cut to a film."

F. Pacing

23. Reveal and return. Success = the 30 seconds opens on the face (we meet the person), roams the B-roll (proof and variety), and returns to the face for one line whose power is the expression. Your justification should name why that line: "I kept 'and that scares me' on the face because the fear is in her eyes; B-roll would have thrown it away."

24. Fix the sizzle reel. Three problems with a 90-second face-free testimonial: (1) we never meet the person, so we don't trust or bond with them — it feels like a voiceover ad, not a testimony; (2) the emotional lines have no face to land on, so nothing hits; (3) it reads as slick and impersonal, the opposite of a testimonial's job. Fixes: open on the face, return to it for the emotional beat and the button, and reserve wall-to-wall B-roll for the informational stretches.

25. Give it a breath. The exhale — a 2-second silent or near-silent beat — makes the moment after it land harder, because the viewer's attention resets and leans back in. Without it, everything runs together and the peak has nowhere to sit. Model observation: "After I added a two-second hold on a quiet cutaway before the final line, the final line suddenly felt like a conclusion instead of just the next clip."

26. Read the Sequence. A model 7-shot table reveals on the face, roams 2–3 matched B-roll shots, returns to the face for a hard line held through a pause, exhales on a quiet beat, and buttons on the face. The grading criterion: are the two most important lines (the emotional one and the button) on the face, and is there at least one held breath? (See FIGURE 30.6 and CS2.7 for the model shape.)

G. Fix the Shot / Fix the Cut

27. Fix the hostage video. Root cause: the interview was cut but never covered — there's no B-roll, so every trim is a naked jump cut. Fix in the edit: lay matched B-roll over every seam (or, if it were two-camera, cut A↔B). But the real prevention was on the shoot — Chapter 20's coverage day. You can't cover seams with B-roll you never shot. (This is the chapter's central ✂️ In the Edit lesson.)

28. Fix the buried spine. It passes sound-off (pretty pictures) and fails eyes-closed (no story for the ear) — so the radio-edit rung broke: the editor cut pictures first and never built a spine that holds as audio. Fix: go back to the transcript, build a real spine, test it eyes-closed, then re-cover. Do not add more B-roll — that's treating the symptom. (FIGURE 30.7: walk back down the ladder to find the broken rung.)

29. Fix the pretty-shot hijack. The skipped discipline is the radio edit — deciding the story in words first, with pictures hidden, so no single shot can hijack the structure. What they do now: set the beautiful shot aside, radio-edit the story properly, and then see whether the shot serves the spine they built (use it) or not (save it for the reel, Ch.39). The story auditions the shot, not the reverse.

30. Fix the drag. The problem is the spine (the words and their order), because it fails the eyes-closed pass. Do not waste time re-cutting B-roll, color, or music — the pictures already work (it passes sound-off). Go back to the radio edit: is the order wrong, is a beat missing, are two bites redundant, is the hook weak? Fix the audio story, then re-check. The most common time-sink is polishing pictures on a broken spine.

H. Settings and scenarios

31. Track assignment. Interview picture on V1; B-roll on V2 (above V1, so it covers the picture while the interview audio plays on). Interview audio on A1 (on top of the mix, always clear); wild sound on A2 (under the voice); music, later, on A3 (dipped under both). B-roll goes above so it can cover V1 via the L-cut; the voice stays on top because it's the spine — sound is half the picture and nothing may fight the words.

32. Single vs. two camera. Single-camera + rich B-roll: hide every internal cut by laying matched B-roll over the seam (the L-cut); you're depending on the B-roll, so lean on it. Two-camera + little B-roll: hide internal cuts by cutting A↔B (the two matched, 30°-offset angles) on the trim — the switch of angle reads as a deliberate shot change, not a jump; save your scarce B-roll for proving claims and establishing.

33. The 90-second CEO. Plan: (1) transcribe the eleven minutes; (2) radio-edit — mark the keeper soundbites, cut the corporate mush, and order the strongest 6–9 into an arc (problem → what they did → result/call-to-action); (3) pull selects, build a radio string-out, test eyes-closed; (4) build the spine, cut the interviewer's questions out, tighten honestly; (5) lay B-roll (product, workplace, the result) over every seam; (6) reveal-and-return so we meet the CEO and stay on the face for the one line that matters; (7) fine cut, both passes, lock. The ethical rule you will not break: no frankenbite — you may make the CEO more concise, never make them claim something they didn't mean, however much the client wants the punchier version.

I. Interleaved

34. Interview → selects. A full-sentence answer ("I've been doing this twenty years") becomes a pullable select — you trim it on the breath and drop it into the spine anywhere. A fragment ("Twenty years," answering an unheard question) becomes an unusable clip you can't lift, because it depends on the question the viewer will never hear. This is exactly why Chapter 19 drilled full-sentence answers — the selects bin is where that discipline pays out.

35. B-roll → cover. For three beats you should name specific matched shots (e.g., "hook: hands working, wide→detail"; "stakes: the empty shelf"; "resolution: the finished piece / a customer"). If a beat has no B-roll next to it, that hole means you'll be forced to leave the talking head naked at that seam, or hold a shot too long — a cost decided back on the Ch.20 shoot. Note it honestly.

36. Rule of Six on a cutaway. On a subject's hard line, holding the face usually wins, and emotion (priority #1, 51%) decides it: the feeling lives in the expression, so cutting away to B-roll — however clean — serves the footage, not the audience. Cut away only if the B-roll adds to the feeling (e.g., the thing they're grieving); otherwise hold. (Straight from Ch.29's Rule of Six.)

37. The mute test on the whole film. Report format: "Eyes-closed pass: [holds / drags at X] → fix the spine. Sound-off pass: [shows a story / collapses at Y] → fix the B-roll. Mute-the-music test: [scene still holds / collapses] → if it collapses, the music was a crutch; rebuild the emotion in the spine and pictures, then bring the bed back." A piece that passes all three is ready to lock. Each failure points to a different fix — don't apply the wrong one.

J. Synthesis

38. Teach it back. A model ~200-word explanation makes these moves: (1) the story of an interview lives in the words, not the pictures, because the voice is the spine; (2) so you build the story on paper/audio first — the radio edit — ordering soundbites into an arc with your eyes closed; (3) you test it by listening: if it holds as pure radio, the pictures can only improve it; (4) then you lay B-roll over the seams. Concrete example: "If I cut pictures first, I'll build a lovely film around a pretty shot that says nothing; if I cut words first, I'll have a story that works, and the pretty shot either serves it or goes to the reel." The strongest answers name the failure mode of doing it backwards (the buried spine).

39. Lock it. Guidance: your two-sentence report should show the passes caught real things. Model: "The eyes-closed pass caught that my middle sagged — two bites made the same point, so I cut one and the story tightened. The sound-off pass caught a 6-second naked talking head with no B-roll, so I added a cutaway; now a muted viewer sees a film." If both passes caught nothing, you probably didn't run them honestly — they almost always catch something on a first lock.

If your cut passes the eyes-closed test and the sound-off test, and you can point to the seam every B-roll shot covers and swear you built no frankenbite, you have edited an interview-driven piece to a professional standard. Project 2 is locked. On to Chapter 31 and the finish.

Chapter 31

Chapter 31 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises. Students should attempt each before reading. Scope values quoted are illustrative.

1. Spot the cast. A deliberate grade is usually consistent across a whole scene, motivated by story/setting, and applied to a corrected base (skin still reads as believable skin even inside the warmth). An accidental cast tends to be inconsistent shot-to-shot, unmotivated, and pushes skin off-natural (greenish under fluorescents, blue in shade). Tell: look at the whites and the skin. A stylized warm scene still has a believable person in it; a cast makes the person look wrong. If a cut to a new angle suddenly shifts the wall's color, that's an uncorrected cast, not a look.

2. The adaptation experiment. After staring, the white wall shows an afterimage in the complementary color (stare at blue → see orange/yellow). This proves the eye actively re-baselines: it adapted to the cast, redefined "neutral" relative to it, and the aftereffect is that recalibration snapping back. Because your eye changes its own reference in under a minute, it cannot be a stable instrument for judging color — the scope, which never adapts, must be.

3. Match-cut hunting. In a well-finished scene the two angles are indistinguishable in wall color, brightness, and skin — the correction matched them. Where it fails (common in low-budget or fast-turnaround work, and multi-camera livestreams), you'll catch the wall warming/cooling or the face brightening on the cut. The trained observation: matching failures are most visible on neutral surfaces (walls) and skin, and at the cut — which is exactly where §31.5 tells you to look.

5. Read the waveform. Two problems, one of them fine: the trace welded flat to 0% means crushed shadows — the darkest detail is clipped and gone (the fix must happen on the shoot, or by lifting slightly if not truly clipped). Topping out at ~85% is not a problem in itself if nothing in frame is meant to be pure white — but if there's a genuine white, raise gain until it reaches ~100%. One fix: bring lift up until the shadow trace lifts off the 0% floor and detail returns (only possible if it wasn't hard-clipped at capture).

6. Read the parade. Blue sitting lower than R and G at both ends of the trace on a neutral wall = the image is short on blue = a warm (yellow/orange) cast. Correction: temperature slider toward cool, or raise the blue channel / lower red-green, until all three line up on the wall. Confirm on the vectorscope that the trace pulls back toward center.

7. Read the vectorscope. Two problems: (1) the whole trace pushed toward cyan = an overall cyan cast (fix first, globally — temperature warmer and/or tint, until the trace re-centers); (2) skin sitting clockwise of the skin-tone line toward yellow = skin too yellow/jaundiced. Fix order: neutralize the global cyan cast first (it's affecting skin too), then re-check skin and nudge it back onto the line. Always fix the global cast before the local skin, because the cast is part of why skin is off.

8. Settings drill: three casts. (a) Green above R and B on a white shirt → green cast → tint toward magenta. (b) Red high, blue low at the top of the parade → warm cast in the highlights → temperature cooler (and check it's not just the highlights — if only the top, it may be a mixed-source issue). (c) The trap: blue high at the bottom only, level at the top, means the shadows are blue but the highlights are neutral — a global temperature move would wrongly cool the highlights. Fix it where it lives: pull blue down in the lift/shadows only (a tonal-zone correction), leaving gamma/gain alone. Lesson: read where in the range the cast sits and correct that zone.

9. Build a scope cheat-sheet (model). Waveform: brightness 0–100%; neutral = spans 0–100 with nothing on a rail; crushed = piled on 0%, blown = piled on 100%, milky = floating off 0%. Parade: R/G/B balance; neutral = channels level on neutral areas; cast = a channel high/low; note where (top/bottom) it's off. Vectorscope: hue=angle, saturation=distance; neutral = trace near center; cast = trace off-center toward a hue; skin = arm on the skin-tone line (~11 o'clock), off-line = wrong skin hue.

11. Set the range. Setting black to 0% and white to ~100% on the waveform typically transforms a flat clip: fog lifts, contrast appears, the image "pops" honestly. Watch for: the trace touching but not welding to either rail (if it welds, you've gone too far — back off). Common surprise: the picture looked "fine" before, but was actually grey and low-contrast; the scope revealed the unused range at both ends.

12. Neutralize a real cast. For an auto-WB indoor shot, the fix is usually a temperature move (warm/cool) plus a small tint (for any LED green); the eyedropper on a neutral reference does most of it in one click. Which slider "did the work" tells you the cast's nature: mostly temperature = a warm/cool source mismatch; mostly tint = an LED/fluorescent green. Confirm on the parade that your neutral reference now reads level R/G/B.

13. Five Ways: one shot, five corrections. (a) neutral is the reference. (b) milky blacks (lift too high): the waveform never reaches 0%, image looks foggy and low-contrast. (c) crushed: trace welded to 0%, shadow detail gone to black. (d) blown: trace welded to 100%, highlight detail gone to white. (e) deliberate blue cast: parade shows blue high, vectorscope off-center — useful to see how obvious a cast is on the scopes even when your eye starts to accept it. The exercise trains you to recognize each failure on the scope instantly.

14. Correct the mixed-light scene (model). There is no single correct white balance for two source colors, so you choose: balance for the dominant, story-important light — the light on the face, usually the window in the Café Scene. Get skin onto the skin-tone line and the face reading naturally; deliberately let the warm lamp stay warm, because that warmth is motivated (a real lamp we can see) and reads as believable. Justification to write: "I balanced the face because that's what the viewer watches and judges; the lamp's warmth is real and motivated, so keeping it makes the room honest, not wrong." (That residual warmth is also the seed of the Ch.32 grade.)

15. Two stills, one look. Put both waveforms up and match the second's black and white points to the first (lift and gain). The insight most students report: a large share of what looked like a "different color" between the two shots was actually just a brightness/contrast difference — once tone matched, the color was much closer than expected. This is exactly why §31.5 says match tone before color.

17. Match to a hero. Correcting the hero, then matching two shots to it: most students find the skin step the fiddliest, because it's the most perceptually sensitive and the last 2% of a match lives there. Tone matches fast (two numbers on the waveform), color matches with the parade, but getting two faces to sit at the exact same point on the skin-tone line takes small, careful nudges. Time it; the routine gets fast with reps.

18. The rainbow event (model). Three lights (window/lamp/overhead) on one subject: pick the best-lit shot as the hero, correct it to neutral, then match the other two — tone first (equalize black/white points), then color (each source has a different cast: window cool, lamp warm, overhead often green — neutralize each on the parade), then skin (all three faces to the same point on the skin-tone line). Result: three obviously different lights become one consistent scene. This is the multicam/event matching problem, and the discipline scales directly to a real three-camera shoot.

19. Fix the crush. The "rich cinematic blacks" are crushed — the waveform trace is welded to 0%, so all shadow detail (the texture in the dark side of the face, folds in dark clothing) is clipped to featureless black and deleted. Fix: raise lift until the shadow trace lifts off the 0% floor and detail returns — but only recoverable if it wasn't hard-clipped at capture; if it was, it's gone. What was lost: every gradation in the shadows, which is where mood and dimension live. Deep blacks are good; crushed blacks are missing information.

20. Fix the order. They graded before correcting. A look is a fixed shift from neutral; the shots were not at the same neutral starting point (different casts, exposures), so the identical look landed differently on each — fine on the near-neutral shots, orange on the already-warm ones. Correct order: (1) correct every shot to a matched neutral baseline; (2) then apply the single look — now it lands identically on all of them because they all started from the same place. "Correct before grade" isn't a preference; it's what makes a consistent grade possible.

21. Fix the sterile café. They misunderstood that "neutral" means every source rendered as grey. It doesn't. Correction balances the dominant/story light (the face) so the person is honest; it does not erase motivated color from practicals we can see. By neutralizing the warm lamps to grey, they stripped the room of its real, believable character and made it feel like a morgue. They should have balanced the face and kept the lamps' warmth — honest, not sterile.

22. Fix the banding (model). Root cause: the shot was pushed too hard for the data it contains — most likely 8-bit, heavily compressed capture (Chapter 3). Stretching contrast and saturation spreads too few tonal steps across the smooth sky gradient, so the steps become visible bands. More correction can't fix it: the missing intermediate values were never recorded; you can only make the banding less obvious (gentler moves, a touch of noise/dither), not truly remove it. Prevention on set: shoot 10-bit (more tonal steps = headroom to push), a higher bitrate, and expose well so the sky doesn't need heavy lifting.

23. Write the workflow. One shot: (1) normalize/set contrast — black to 0%, white to 100%, no crush/blow; (2) neutralize cast on the parade; (3) check skin on the vectorscope; (4) sanity-check. Scene: (5) pick hero, match rest (tone→color→skin); (6) play at speed, fix jumps. Step 1 before step 2 because a wrong brightness/contrast makes color hard to judge (and a too-dark shot looks like it has a cast). Matching (5) comes last because you can only match shots to a hero once the hero is fully, correctly corrected.

24. The monitor problem. The uncalibrated laptop screen and the colored room light shifted what the editor saw, so they "corrected" to compensate for their screen, not the file — and on a different (also imperfect, but differently imperfect) phone screen, the real state of the file showed through as wrong. The single free fix: trust the scopes. They report the file, not the screen, so a black point set to 0% on the waveform is 0% on every device regardless of what your laptop shows. (Working in a neutral, unchanging room light helps too, and costs nothing.)

25. The on-set fix (model). Four things before rolling a two-camera interview: (1) lock both cameras to the same manual white balance (not auto) → they start color-matched and don't drift between takes; (2) match picture profiles on both → you're not reconciling a Log camera with a standard one; (3) shoot a grey card under the light on each camera → one-click neutral balance and a shared reference to match to; (4) expose both to matched brightness on the waveform → tone matches, so post is a nudge not a rescue. Each directly shortens the correction pass and makes the shots match closer out of the gate.

27. Match a Project 2 scene. "Done" looks like: play the scene at full speed and no cut makes the wall, room, or skin visibly change color or brightness. On the scopes, every shot's black/white points sit together, the parades agree on neutral surfaces, and every face rides the same point on the skin-tone line. If any cut still jumps, that shot isn't matched — return to it. This is real, gradeable progress on Project 2's finish.

29. Exposure to correction. The correctly-exposed clip corrects easily — clean shadows, honest color, gentle moves. The under-exposed clip fights you: lifting it to match amplifies noise (Chapter 5), the shadows crawl, and color gets unreliable in the dark areas. Two sentences: On-set exposure decides how much clean information correction has to work with; a well-exposed shot corrects with a nudge, an under-exposed one only stretches thin, noisy data. "Fix the exposure in post" is a lie about shadows — you can brighten them, but you reveal the noise the camera recorded instead of the detail it didn't.

30. The whole pipeline in one scene (model). A strong answer traces a real scene through: the frame/shot sizes chosen (Ch.6–7) that gave the editor coverage; the light (Ch.11–13) that set the mood and the casts correction now handles; the audio captured (Ch.14–15) awaiting the mix; the cuts (Ch.26–28) that assembled it; and now correction pulling it all to a matched baseline. The most helpful on-set decision is usually locked manual white balance + a grey card + good exposure; the most harmful is usually auto white balance drifting between takes or under-/over-exposure that limited how far color could go. The point: the color page inherits every on-set choice — you shoot for the color pass.

Chapter 32 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered exercises. Many items are open-ended grading/seeing work; for those the "answer" is a rubric of what a strong response demonstrates. Try each before reading. (Numbering matches exercises.md.)

A. Seeing color

1. Name the look. (Answer.) Strong answers pair a palette phrase with a mood and a motivation: a prestige drama's "desaturated steel-blue = cold, controlled dread, because the show is about institutional coldness"; a travel vlog's "warm gold = aspiration and escape." The lesson: a look is always saying something, and if you can't name what, it may be unmotivated — or you missed the story reason.

2. Spot the teal-orange. (Answer.) You'll lose count — it's the era's default. Used well: restrained, motivated, skin held on the flesh line, you don't notice the color before the content. Overdone: shadows visibly cyan, skin plastic-orange, every frame the same two colors. The reliable tell is whether you notice the color before the content.

3. Where does your eye go? (Guidance.) The eye lands on the brightest, most saturated, most in-focus thing. Strong answers spot the subtle "subject slightly brighter / background slightly darker" treatment — a reverse-engineered power-window vignette — and realize attention was directed, not left to chance.

4. Find the light (skin check). (Answer.) "Off" faces usually lean green (bad white balance or fluorescent spill), orange (an over-warm grade), or waxy (over-smoothed or over-saturated). Naming the direction of the error by eye is exactly what the vectorscope makes objective in §32.5.

5. Correction or grade? (Model.) Correction = it just looks right, matched, neutral; grade = a deliberate mood on top. The insight: in good work the seam is nearly invisible — that's the point. Strong answers point to at least one clearly creative choice (a warm cast, a crushed contrast) as distinct from the "it just looks correct" baseline.

B. Reading graded sequences

6. Picture the pipeline. (Answer.) Source → correction (balance + neutralize) → look (motivated color/contrast) → secondaries (shape parts, protect skin) → output (transform to Rec.709). To check the correction is clean under the style, toggle the look node off — you should drop back to the neutral base.

7. Diagnose the look (Amélie). (Answer.) The three things that make a bold look work: (1) motivated ("a fairy-tale Paris"), (2) consistent across every frame, (3) built around protected skin. Failure briefs: a wedding film pushed heavy teal-orange for no reason (unmotivated); a grade that drifts warm-to-cool shot to shot (inconsistent); a look that turns the couple orange (skin off the line).

8. Read the café grade. (Answer.) Changes: warmed highlights, shadows lifted to warm brown, richer/warmer lamp pools, a gentle saturation bump. Protected: (a) the daylight window — held near-neutral so it still reads as daylight, giving the warm room an honest cool anchor; (b) skin — held on the flesh line so people look healthy. All motivated by "a warm place someone loves."

9. Write your own signature-grade Described Shot. (Model.) A strong seven-field shot puts THE LIGHT and THE EFFECT on the grade (palette, contrast, what the color says), attributes it [after <title>, <year>], and lands a LESSON tying the look to story. Weak answers describe the frame's content but never name a color decision — the whole point.

10. Read Case Study 1's consistency. (Answer.) Holding a look through night scenes is harder because the source light itself pushes cold/blue, fighting a warm look — the grader must actively pull it back into the film's palette rather than let it drift. A "lazy" grade lets night go cold-blue while day stays warm, and the film fractures into two worlds instead of one. A look that holds noon-to-midnight has a spine.

C. Building looks

11. Correct, then look. (Guidance.) Toggling the look node reveals the difference between "right" (correction) and "feels like something" (grade). If adding the look makes the shot worse or fights you, the correction underneath wasn't clean — fix the base first.

12. One clip, three moods. (Trade-offs.) Warm = comfort/nostalgia; cool = tension/isolation; desaturated = bleak/grim. The "best" is whichever matches what the clip is about — the real lesson is that the correct grade is content-dependent, not a default you apply to everything.

13. Build the warm café look. (Model.) Mirrors Case Study 2: a warm primary look node, a skin qualifier holding faces on the line, a tracked window taming the hot glass, a node richening the lamp pools, a soft vignette to the subject, then consistency + a 20% pull-back. Success: the room reads warm and every face stays believable on the vectorscope.

14. Name it in a sentence first. (Guidance.) If the viewer's felt-word matches your one-sentence intent, the grade communicated. A mismatch means it was either unmotivated (said nothing) or overdone (said the wrong, louder thing). The sentence is the test.

15. Grade a whole scene for consistency. (Model.) Correct each shot individually to neutral, apply one shared look node across all, watch the scene through at speed, and fix any jump on the individual correction (not by re-doing the look). Strong work treats the look as a property of the piece, not one hero shot.

16. Amplify, don't impose. (Guidance.) Reading the scopes first, students find the image already leans a direction; a look that amplifies that lean (warming an already-warm room) feels like it belongs to the footage, while an imposed color (cooling that same room hard) feels applied and fights the light. The lesson mirrors motivated lighting: amplify a real tendency rather than inventing one.

D. Secondaries and skin

17. Draw a vignette. (Guidance.) Soft, wide-feathered, tracked if the subject moves; invisible as an effect while still pulling the eye. If a friend can see a dark oval, it's too strong — feather wider and reduce it.

18. Qualify the sky (or a wall). (Answer.) Feather the qualifier's edge until the adjustment is seamless. Over-doing it produces the tell-tale halo/"cut-out" fringe around the selection — exactly what "too far" looks like, and why edges must be soft and moves restrained.

19. Put skin on the scope. (Model.) Both complexions trace the same flesh-tone line at different lengths: the darker complexion closer to center (lower saturation/luma), the fairer further out — same angle, different distance. If either leans off the line, the white balance is off; correct it before grading.

20. Protect skin under a look. (Model.) With a strong warm look, skin drifts toward orange; a skin qualifier pulls it back to the line while the room stays warm. Toggling flips the shot between "a person in a warm room" (on) and "a warm object" (off). This single move is the heart of the chapter.

21. Relight in post. (Answer.) A tracked power window darkens the hot window or lifts the shadowed face; it must track so the effect stays glued as the subject or frame moves. Full credit names the one-sentence motivation ("the window was pulling the eye off the person") and confirms the track holds.

22. The three-secondary discipline. (Model.) Exactly three motivated secondaries (vignette, skin qualifier, one relight/fix), each subtle enough a viewer can't point to it. Toggling all three should make the shot clearer and more focused, not more obviously colored. If it's the latter, everything is too strong.

E. Fix the grade

23. Fix the one-click LUT. (Answer.) In order: (1) disable/remove the LUT; (2) correct the shot — exposure, white balance, black/white point; (3) re-add the LUT on its own node at reduced strength; (4) hand-trim toward what the shot needs; (5) qualify skin and restore it to the flesh line. The LUT is a starting point, not the step.

24. Fix the sick skin. (Answer.) Cause: a global warm grade dragged skin off the flesh line toward orange. Fix: a qualifier node selecting the skin, pulled back onto the line, while the room stays warm. Selective protection — never weaken the whole look to save the faces.

25. Fix the inconsistent scene. (Answer.) The skipped step was matching/correcting each shot to neutral before applying the look (Chapter 31). A single LUT on mismatched shots amplifies the mismatch. Correct order: correct + match every shot to a neutral baseline (waveform/parade/vectorscope, skin on the line), then apply one shared look.

26. Fix the over-graded piece. (Answer.) Diagnose against the four disciplines: unmotivated (reduce toward the brief); no restraint (pull the grade back ~20%); skin off the line (qualify and restore); templated/inconsistent ("like everyone else" = the teal-orange trap; motivate a look specific to this brand). One concrete fix per problem.

F. Settings & pipeline

27. Order the node tree. (Answer.) balance → contrast → LUT/convert → look → secondaries/skin → output to Rec.709. Rationale: correct before you style; LUT/convert between correction and the hand-built look; selective moves after the global look; the delivery transform last.

28. Write the grade plan. (Model.) (a) Café: "cozy and premium" — warm primary + skin qualifier + tame the window. (b) True-crime interview: "cold dread" — cool, desaturated, controlled contrast + skin held so it doesn't go green. (c) Product: "clean and bright" — near-neutral, crisp whites, controlled saturation + honest skin. (d) Wedding: "warm and nostalgic" — gentle gold, soft contrast + skin protected. Each names the sentence, the node order, and the protection.

29. LUT, honestly. (Answer.) Workflow: after correction, own node, reduce strength, check clipping/crushing and check skin on the scope, then hand-trim. Don't use it when it crushes shadows or clips highlights badly, drags skin off the line, or was clearly built for a very different camera/exposure.

30. Spec the deliverable. (Answer.) Check: output color space is Rec.709; spot-check on a second screen or two (a phone and another monitor); re-verify skin on the flesh line; watch the whole piece straight through for consistency; confirm nothing clips. Each matters because the grade is built for the average viewer's device, not your one heroic monitor.

G. Recreate & Five Ways

31. Recreate a famous look. (Guidance.) Analyze the target's palette, contrast, and skin treatment, then build by hand and compare side by side. What's usually "missing" isn't the color — it's the skin discipline and the consistency. Getting close teaches that a look is mostly restraint and protection, not saturation.

32. Five Ways: one face, five looks. (Model.) Neutral, warm, cool, gritty/high-contrast, soft/low-contrast — skin held on the flesh line in all five. Strong work notes how far the world can be pushed while the face stays believable, and assigns each look a home (warm=comfort, cool=tension, gritty=grim, soft=romance/nostalgia).

H. Interleaved & synthesis

33. Grade meets capture (Ch.3). (Answer.) The heavily-compressed, standard-profile clip breaks first: banding in smooth gradients (sky), blocky/noisy shadows, skin falling apart under the push. Capture higher bit depth (10-bit), less chroma subsampling (4:2:2 if available), a higher bitrate/less-compressed codec, and Log for latitude — you can only grade what you captured.

34. Grade meets light (Chs.11–13). (Answer.) Lean into (amplify) the source that carries your mood — for a warm scene, richen the practical — and protect the other as an honest anchor (hold the cool window near-neutral). A motivated grade amplifies light that was really there rather than inventing a color the scene can't justify.

35. The finishing chain (Chs.30–33). (Guidance.) Order: lock → correct → grade → mix. Key mood decisions: the lock sets story shape; correction sets honesty; grading sets the visual mood; mix sets the aural mood. Grading a locked cut saves redoing tracked grades that would come loose if the cut moved.

36. Teach it back. (Model.) A strong ~200-word explanation states: correction = "make it right," grading = "make it feel"; skin is protected above all because the eye is a pre-conscious face-detector; teal-orange is useful (complementary warm/cool separation) but a trap when overdone (templated, skin off the line). Uses one concrete example and lands the thesis that a grade must be motivated, consistent, and skin-safe.

37. The self-audit. (Guidance.) No single answer. Strong work scores motivation / consistency / restraint / skin honestly (1–5), fixes the lowest specifically, and keeps the note as a baseline for the Chapter 39 reel. Grading your own work critically is the real deliverable.

38. The Frame-Log color habit. (Guidance.) No single answer. A strong week shows a shift from naming content to naming color decisions — a look, an off face, a guided eye, an overdone teal-orange. That shift is the colorist's eye developing, and it's the same skill pointed at your own footage.

Chapter 33

Chapter 33 — Answers to Selected Exercises

Model answers and critiques for the starred and odd-numbered exercises (and the even items flagged "Answer provided"). Treat each as one strong response, not the only one; audio judgments are ranges. Numbering matches exercises.md.

1. The volume-knob test. Every volume grab is a leveling or mastering failure. Reaching up means dialogue dropped below a comfortable level (uneven leveling, no compression, or too-loud music/ambience masking it); reaching down means a peak or a music cue was too hot (no ride, no limiter, or an over-loud master). Strong answers map each grab to a §33.3/§33.6 fix — compression + hand rides for the ups, ducking + a master limiter for the downs — and note that most grabs are leveling problems. The lesson: a professional mix never makes the viewer touch the volume.

2. Find the bed. In polished work the bed sits perhaps 10–20 dB under the dialogue and ducks — you can hear it lift in the pauses and bow under each line. If you cannot hear it lift and dip, it is either static (amateur) or so buried it may as well not be there. The point is to start hearing ducking as a deliberate, dynamic move rather than a fixed background.

3. Name the noise. (Diagnostic drill — no single answer.) Listen for: hollow/echoey = room reverb (worst to fix in post); a steady tonal buzz = electrical hum; a broadband "shhhh" under everything = noise floor/hiss; loud-quiet lurching = leveling; crackle/harsh distortion = clipping; can't-hear-the-words-under-the-song = music too loud. Naming it is the whole skill; each maps to a specific §33.2–33.4 tool. Most amateur clips are two or three of these at once.

4. The layers of a place. Even a "simple" café scene is typically 4–6 layers: dialogue, a room-tone/ambience bed, a distant-murmur ambience, two or three spot effects (door, machine, crockery), small Foley (footsteps, cloth), and a music bed. The realization is that "realistic" sound is built from many quiet layers stacked below the dialogue, not captured whole in one take.

5. Sound-off, sound-on, sound-only. With picture only, you get geography and action but weak mood, no offscreen world, and little emotional steering. Sound-only usually delivers more story than picture-only — location (ambience), events (effects), and how to feel (music, silence). The paragraph should conclude that sound is not "added to" the picture; it carries at least half the information, which is the chapter's whole thesis.

6. Picture it. Priority order, loudest→quietest, with level relative to dialogue: Dialogue (the reference, ~-12 dBFS peaks — carries meaning, always clear) → Spot effects (~6 dB under — punctuate actions, synced) → Music bed (~12–18 dB under, ducking further under lines — mood) → Ambience (~18–22 dB under — establishes place, hides edits). Dialogue is the reference everything else is set against.

7. Diagnose the field. The espresso machine does two jobs: (obvious) it gives the counter craft, texture, and rhythm — it puts "coffee" in the scene; (hidden) as a loud, continuous sound it masks a dialogue edit beneath it, because a louder continuous sound covers the small discontinuity of a cut. This is the same masking principle as the waterfall in Case Study 1, where a wall of sound covers the characters' voices.

8. The missing layer. Sample answer: a kitchen order-ticket "ding" or a milk-steaming hiss, sitting at ambience-to-SFX level (~-24 dBFS), synced to a background action, motivated by other customers being served. It deepens the sense that the café is a working place with life beyond the two speakers. Any answer works if it sits below dialogue, is motivated by an event, and adds off-screen life.

9. Write your own build. (Open.) A strong table lists layers in add-order (cleaned dialogue → ambience bed → spot effects → Foley → music), each with a job and a level relative to dialogue (dialogue = 0 reference; ambience -18 to -22; effects -6; music -12 to -18, ducking further). Grade on whether every non-dialogue layer is placed below dialogue and justified by story rather than habit.

10. Kill the rumble. Sweeping the high-pass up, most voices stay untouched through ~80 Hz and only start to thin (losing chest and weight) somewhere around 100–150 Hz; a deep voice wants the cut nearer 70 Hz. The takeaway: ~80 Hz removes rumble that carries no voice; go higher only until the voice thins, then back off. Rumble and hum are separate problems — the high-pass removes broadband low-end energy, but a tonal hum needs a de-hum notch (Exercise 11's cousin).

11. Learn a noise print. At ~6 dB the reduction is subtle and totally clean; at ~10 dB the drone is gone and the voice is still natural — usually the sweet spot; at ~20 dB the voice turns watery, swirly, "underwater" as the tool chews into the voice itself. Keep the ~10 dB version and note the value. The lesson: noise reduction has a cost, and "just enough" beats "all gone" every time.

12. The full cleanup pass. After spot-fix + high-pass + de-hum + modest NR, expect the clip to be dramatically cleaner — no rumble, no hum, a much lower floor — but any room echo will remain (post reduces it poorly), and any distortion/clipping is permanent. Naming what survived cleanup is the point: it teaches which problems must be solved on set (Chapters 14–15), not at the timeline.

13. Room-tone patch. Lay a bar of room tone across the "blinking" cut, crossfade its edges, and the background becomes continuous — the cut vanishes. Before/after should be night-and-day: a cut you could hear becomes one you cannot. This is the cheapest, highest-impact single trick in dialogue editing, and it is why Chapter 15 made you record room tone.

14. The rescue challenge. After the full §33.2 chain, you can lower the fan drone and rumble and even out levels — but the room echo will defeat you: de-reverb reduces it slightly and then starts making the voice sound artificial. That wall is the lesson. Post can clean noise and shape tone; it cannot un-smear reverb or un-clip distortion. This is the visceral argument for close-micing and taming the room on set.

15. Subtractive first. Cutting a couple of dB at the boomy 250–400 Hz region clears the mud directly and adds nothing; boosting the highs to "compensate" would add harshness and loudness without removing the underlying boom, and pile energy into an already-busy signal. In one sentence: cutting the problem is cleaner than boosting around it.

16. Presence and air. A +3 dB lift at 3–5 kHz makes the voice clearer and more forward; +1–2 dB at 10 kHz adds sheen. Overdone, the same moves turn the voice harsh and spitty (sibilance), at which point a de-esser on ~5–8 kHz restores comfort. The lesson: presence and air are powerful and easy to overdo — small moves, and de-ess after boosting.

17. Even it out. Hand-drawn clip-gain automation is the most transparent (you fix only the specific swing, nothing else moves); a gentle compressor (2.5:1, a few dB) handles continuous small variation automatically but can dull dynamics if pushed. Use automation for big, obvious swings and a light compressor underneath for the rest — both, not either. Level the result to ~-12 dBFS peaks.

18. Match two shots. Bring the boomy/dull one down in the mud and up in presence; add body to the thin/bright one and match its presence; then match their overall levels. Cut rapidly between them — success is when the join is inaudible, so the two recordings read as one room and one conversation. This is the audio cousin of shot-matching (Chapter 31).

19. License audit. A correct audit tags each track with a provable right: a library license (with receipt), a Creative Commons license (with the exact terms met — attribution/non-commercial), original/commissioned, or a platform-cleared library. Anything you cannot justify is flagged red to replace before publishing. "I found it online" and "I credited them" are not licenses.

20. Pick the bed. Judge each candidate on (a) energy/arc fit — does it build and rest where the story does? and (b) does it leave the presence range (3–5 kHz) clear for the voice? The winner is usually the warmer, sparser track, not the most exciting one — because the bed's job is to support, not to star.

21. Duck it. With manual automation, draw the music down ~6–10 dB under each line and back up in the gaps; with a sidechain, the dialogue triggers the duck automatically. Note the level difference: music might sit ~-24 dBFS in a gap and duck to ~-32 dBFS under a line. Sidechain is faster and consistent; manual is more controllable on tricky material. Success = the music breathes around the voice.

22. One theme, three moods. Deploy the same cue differently: the opening uses its full, bright arrangement up in the mix; the low point uses a stripped, quieter section (or just its pad) ducked far down; the ending brings the theme back, resolved, swelling in a dialogue gap. The recurring melody binds the piece; the deployment changes the mood. This is the Up lesson (Chapter 1) on your own timeline.

23. Recreate it (a famous cue move). The move to reproduce is the coincidence of a musical change and an edit: place your licensed cue so its beat drop, swell, or downbeat lands on the exact frame of a cut or a story realization. Self-review: toggle the music off and the cut should feel ordinary; toggle it on and the cut should feel inevitable. That felt inevitability — picture and music changing together — is the whole point, and it costs nothing but timing.

24. Five ambiences, one shot. The same shot reads as a different place with each bed — the picture is neutral, the sound assigns the location. Usually the most specific, least-generic bed reads clearest (a distinct café murmur beats a vague "room noise"). The lesson: ambience is not filler; it tells the viewer where they are, and specificity sells it.

25. Record your own SFX. (Practice.) Synced correctly, the effects disappear into the reality of the scene. The deliberately-offset one is the teacher: even a third-of-a-second error reads as clearly wrong, because the eye and ear disagree. Most people find the mismatch obvious by ~1–2 frames of offset for a sharp transient. Sync is frame-exact, not "close."

26. The three-layer soundscape. Model self-critique: mute all three added layers and the scene is "a void with talking"; unmute and it becomes a place. If it does not transform, the ambience is likely too quiet (raise it toward -28 to -30 dBFS) or the spot effects are off-sync (nudge them to the exact action frame). Every added layer must sit below the dialogue.

27. Design a place that isn't there. (Open.) The storm cabin = a low wind-and-rain ambience bed, occasional thunder spot effects, a creak or two of timber, maybe rain-on-window Foley — all below the dialogue. The subway platform = a rumbling-crowd ambience, a distant announcement, an approaching-train swell and brake screech synced to nothing on screen. Same picture, two worlds — proof that sound builds what the picture only implies.

28. Read the two meters. The true-peak/dBFS meter tells you about clipping (keep the max below -1 dBFS); the integrated LUFS meter tells you how loud the program feels (set it to the destination target, ~-14 for streaming). They are different measurements and you need both — level by peaks, master by loudness.

29. Master to spec. The same mix needs a different master level for each destination — you raise or lower the overall level until the integrated reading hits -14 (streaming) or -16 (podcast), while the -1 dBFS true-peak limiter guards clipping in every case. The mix balance does not change; only the final master level does. This is why you keep the mix and master as separate steps.

30. The over-limiting trap. When both versions are normalized to the same playback loudness, the -14 LUFS master sounds better — more dynamic and alive — while the crushed -8 LUFS version sounds flat and lifeless at the same volume. This proves the "loudness war" is unwinnable where platforms normalize: over-limiting costs you dynamics and buys no loudness.

31. The four-system QC. Likely findings: headphones — everything sounds fine (the trap); laptop — bass-light, dialogue mostly OK; phone — thin dialogue if it leaned on low end, muddy if the music is too loud; car/TV — reveals real-world balance and any harshness; mono — stereo-spread effects or doubled sounds partly cancel (phase). Fix each and re-check. A mix that survives all five is a real mix.

32. Fix the described mix. In order: (1) distorts on phone but fine on headphones + mastered to -8 LUFS → the master is too hot/over-limited and low-end-heavy; re-master to ~-14 LUFS with a -1 dBFS true-peak limiter and high-pass any low rumble. (2) jumps loud-to-quiet → level with compression plus hand rides to a steady ~-12 dBFS. (3) no background, cuts click into silence → lay a continuous ambience/room-tone bed. (4) song at nearly dialogue level → duck the music well under the voice, and confirm the song is licensed (a pop song usually is not). Fix master/level/ambience/music, then QC on phone + mono.

33. Fix the plan. Legal problems: a "trending song" added in-app is almost certainly unlicensed for that use and will be muted/blocked/demonetized — credit and non-profit are not defenses. Craft problems: finishing audio after upload means no cleanup, leveling, ambience, or loudness mastering — the exact half of "professional" this chapter is about — and doing it "in the app" gives none of the tools. Instead: mix and master before export (Chapter 36), using licensed music.

34. Shoot for the mix. Five, each mapped to a post step: record 30–60 s of room tone → the noise print + ambience seed (§33.2/33.5); mic close and on-axis → less noise to reduce, clearer presence (§33.2/33.3); set gain for ~-12 dBFS peaks with headroom → nothing to un-clip (§33.3); tame the room / avoid reverb → the one thing post can't fix (§33.2); record clean wild SFX like the door and machine → your own spot effects (§33.5). Every on-set discipline is a post shortcut.

35. The finish, in order. Correct order: color correct → color grade → noise-reduce dialogue → add music bed → add lower thirds → master to LUFS. Picture finishing (correct then grade, Chapters 31–32) and audio finishing can proceed in parallel, but within audio: clean dialogue before adding music; graphics (Chapter 34) are their own layer; and mastering to LUFS is always last, on the finished mix — mastering first would be meaningless, because there is nothing final to measure.

36. Teach it back. Model (≈200 words): "Streaming platforms don't play your file at the volume you exported — they measure how loud it feels, in LUFS, and turn it up or down to hit their own house target, around -14 LUFS. So if you crush your mix with a limiter to be as loud as possible — say -8 LUFS — the platform just turns it down to -14 to match everyone else. You gain no loudness at all. But you've paid a price: to get that loud you flattened your dynamics, so now your mix plays at the same volume as a well-mastered one and sounds worse — smaller, more fatiguing, less alive. Mastering 'to the target' means you set your integrated loudness at the platform's number (-14 for streaming) and stop, keeping your true peaks under -1 dBFS so nothing distorts. Your quiet parts stay quiet and your loud parts stay loud — the dynamics that make it feel real. The old 'loudness war,' where everyone competed to be loudest, is simply unwinnable once playback is normalized. So don't fight for loudness the platform will erase; master to the target and keep your dynamics."

37. The full finish on Project 3. (Guidance.) Run the CS-02 session order on Project 3: stems → clean/level dialogue → ambience → synced SFX → licensed ducked music → mix and master to ~-14 LUFS, -1 dBFS ceiling, QC on phone + mono. The three sentences on "what sound added that the grade could not" should name things color cannot do: place (ambience made it somewhere), clarity/trust (clean, level dialogue), and feeling/steering (the ducked music). If the reader cannot feel a difference, the mix is under-built — usually the ambience is too quiet or the dialogue was never leveled.

Chapter 34 — Answers to Selected Exercises

1 (The lower-third audit). Strong observations note that pros keep name supers off the eyes and mouth, hold them roughly four to six seconds, and usually introduce a name once. If a broadcast re-supers on every cut, that's worth noticing — rolling news does it because viewers join mid-broadcast, whereas documentary shows a name once. The lesson: the "show it once" rule flexes by genre and context, but the placement and read-twice rules hold everywhere.

3 (Information or noise). Grade on whether each graphic is assigned a real job: a location super = orient; a stat = clarify; a persistent corner logo = brand (borderline — justify it); a random animated swoosh between shots = noise. The skill being built is the willingness to label something "noise" and cut it. A student who finds every graphic "fine" isn't looking hard enough.

5 (The sound-off test). A model critique names what worked for the muted viewer (key words emphasized, text synced to the speech, a readable pace, strong contrast) and what failed (a line that flashed too fast to read; low-contrast text over bright footage; a joke or fact that lived only in the audio). The win is auditing specifically for the muted viewer, who is often the majority on social.

7 (Diagnose the field — name on the first frame). With the name on the very first frame, THE CUT becomes an instant slam-up the moment we see the person, and THE EFFECT shifts to "we're labeled at, like a news chyron, before we care who this is." The delayed version (six seconds in) lets us settle on the face and start to wonder who they are, so the name arrives as the answer to a question we've begun asking. Lesson: the timing of a graphic is part of its meaning, not a detail.

8 (Re-time the kinetic quote). Whole-quote-at-once with only the final word animating is better when the line is short and you want the payoff to land cleanly — it's calmer and more elegant. Building phrase-by-phrase is better for a longer line, a spoken read you're tracking word by word, or when you want rhythmic energy and sync. Trade-off: the build buys energy and synchronization; the all-at-once buys calm and single-word emphasis. Neither is "right" — they serve different lines.

9 (Write your own Described Sequence). Grade on whether the table captures each element, its on-screen action, the sound under it, and how it leaves — and whether the paragraph explains why the graphics are timed as they are (e.g., "each label lands on the narrator's emphasis" or "the title holds through the establishing shot, then clears for the first line"). Good answers name sync and restraint; weak ones just inventory what appeared.

11 (One title, three voices). The serif reads editorial/heritage/documentary; the bold sans reads modern/product/tech; the damaged display reads horror/edgy/underground. The point is that the same three words mean differently in different type — the letterform is content, not decoration. Reward students who can put the feeling each font evokes into words; that articulation is the skill (§34.2).

13 (The hierarchy fix). A model redesign makes the title largest and boldest (it's the name of the thing), the subtitle medium weight (it supports), and the date smallest and lightest (it's metadata). The one-sentence justifications should tie size and weight to importance: "the title is biggest because it's what the viewer most needs; the date is smallest because it matters least." The lesson: size and weight alone create a reading order — no color or motion required.

15 (Put it in the dead space). If the framing allows, the super lands in the subject's lead-room third, off the face, inside title-safe, with a subtle scrim or bar for separation. If the shot is centered with no dead space, the win is noticing the problem: either reframe next time (a shoot-for-the-edit lesson) or super the name over a cutaway/B-roll of that person. Recognizing "I have nowhere clean to put this" is itself the learning — it traces back to a framing decision (Ch.6, Ch.19).

17 (Five Ways: one name). No single right answer. Typically the version with lead-room placement, a subtle scrim, a medium ease, two lines, and no sound reads as the most "designed." Three lines and a sound sting usually over-build a calm piece. Reward reasoning about restraint over any particular pick — the student who can say why their choice is quietest-that-works has the lesson.

18 (The recurring-name test). Re-showing each name on every cut back feels like an automated news chyron and quietly nags the viewer; showing it once feels authored and confident. The lesson: introduce a name once, a beat after the person starts, unless the piece is long enough that a much-later reminder is warranted. If the student prefers the repeated version, ask them to watch a well-made documentary and count how often it supers a name (answer: once).

19 (Two keyframes, one fade). At 24–30 fps, a 15-frame fade means the software interpolated roughly thirteen or fourteen in-between frames from your two keyframes — you set the endpoints, it drew the middle. The whole point is to see interpolation happen and to internalize that animation is set at the keys, not drawn frame by frame.

20 (Linear vs eased, back to back). The target sentence: "The eased version decelerates into place, so it feels like a real, weighted object settling; the linear version travels at constant speed and feels mechanical and cheap." That single distinction is the amateur-versus-professional line for almost all motion (§34.4). If a student can't feel the difference yet, exaggerate the move (a longer slide) until they can.

21 (The overshoot). A tiny overshoot adds life to energetic or playful pieces (a snappy brand, a kids' show, a hype reel); it feels gimmicky in somber or serious ones (a documentary, a memorial, a corporate testimonial). The restraint rule: a small overshoot reads as craft, a big bounce reads as amateur. Match the amount to the tone — and when in doubt, leave it out.

22 (Match the pacing). A model answer lands on two different frame counts — say, an 8–10-frame snappy move for the fast-cut clip and a 16–20-frame gentle move for the slow, contemplative one — and explains that the motion has to belong to the rhythm around it. A bouncy fast super in a slow elegy feels as wrong as hard rock under a funeral. Reward the reasoning, not the exact numbers.

23 (The count-up). The number should ease onto its final value (decelerate as it lands), which is what makes data feel earned rather than merely printed. Good placements in Project 3: over B-roll at the moment a stat is mentioned, or on the end card. A label beneath keeps it meaningful for a muted viewer (§34.5, CS-02).

24 (Sync to the beat). On the beat, the type feels composed, intentional, and musical; knocked a few frames off, it feels loose and subtly "wrong" even when the viewer can't name why. The lesson is concrete: lay the sound first, find the beats or stressed words, and place your keyframes on them. Sync is most of what separates good kinetic type from noise.

25 (The underline wipe). A single eased rectangle wiping in beneath a title (two keyframes on its width or position) adds a "designed" feel for almost no effort, and it's endlessly reusable — it belongs in your style kit. The lesson: simple animated elements are two or three keyframes and an ease, not complex effects.

26 (The 10-second explainer). Grade on: one idea only; kinetic type plus one animated element (a stat, an arrow, a map dot); everything held to a single style kit; changes landing on a beat; and — critically — all text readable at the speed it moves. Strong answers stop at one element and sync to sound; weak ones cram three ideas and five effects into ten seconds.

27 (Photosensitivity check). Pass/fail: no element flashes more than three times per second, and there's no large, high-contrast full-frame strobe. If an element fails, slow it or soften the contrast. This is a non-negotiable accessibility gate, not a stylistic preference (§34.5).

28 (Fix the failed lower third). Sort the errors by category. DESIGN: thin grey script over footage → a clear font with separation (scrim or stroke); four lines → cut to two (name + role). PLACEMENT: centered across the mouth → move to the lead-room dead space, off the face. TIMING: appears on the first frame → delay a beat or two; holds 12 seconds → hold only read-twice; re-appears on every cut → show once. MOTION: constant-speed slide → ease it (ease-out to land). Naming the category for each problem is the skill being tested.

29 (Fix the title). Fixes: four fonts → one or two chosen for voice; thin white over a bright drifting sky → add a scrim or stroke and a heavier weight, and scrub the whole shot to confirm; text touching the frame edge → pull inside title-safe (~90%); on screen one second → hold long enough to read aloud twice. The tie that binds them: legibility plus restraint — the two governing ideas of §34.2.

30 (Fix the over-animated open). It fails because every choice — the spin, the per-letter bounce, the lens-flare template, the whoosh — is decoration that actively fights a somber tone about loss; it violates motivate every choice and restraint. A redesign in words: a slow, quiet fade-up of the title in a warm editorial serif over a gentle opening image, no sound sting, inside title-safe, held to be read twice. The single principle: the graphics must serve the story's feeling, and this story's feeling is quiet.

31 (Write the style kit). Model instincts. (a) Bank testimonial: a clean corporate sans, the brand blue, conservative sizes, an understated ease, no sound sting — trust and calm. (b) 9:16 product reel: a bold sans, a punchy accent color, larger text kept well inside the tight social safe zones, a snappy ease, maybe a small sting — energy and legibility while muted. (c) Documentary short: an editorial serif paired with a clean sans, a muted palette, gentle slow eases, no sting — warmth and restraint. Reward answers that match the kit to the tone and brand, not "correct" fonts.

32 (Spec the safe zones). In 16:9: all text inside ~90% title-safe, the lower third sitting above the bottom safe line. In 9:16: everything moves further in, because the platform's interface (captions, buttons, username, description) overlaps the bottom ~15–20% and sometimes the top — so supers and captions belong in the central band, higher than in 16:9, and never at the very edges. The reason is simply that on vertical, the app's UI covers the frame's margins.

33 (Layer it on the timeline). Each graphic goes on its own video track above the footage (picture on V1, graphics on V2–V4), timed to the story beats. The confirmation step is the whole lesson: turn the graphics tracks off and the film should still play and make sense — proving each graphic was an information layer, not a crutch propping up the edit (§34.1, Ch.26).

34 (Interleave — the finished 30 seconds). Grade on all four layers being present and coherent: a J- or L-cut where the sound leads the picture (Ch.28), a neutral color correction (Ch.31), a music bed under the dialogue (Ch.33), and a title plus a lower third (this chapter). Usually the hardest part is timing the graphics to the cut and the mix so they feel of a piece. The meta-lesson: finishing is a stack, and each layer assumes the one beneath it is done — which is why this chapter comes after color and sound.

35 (Interleave — frame for the graphic). The reframed shot should leave a genuinely empty lead-room third; compared to a centered original, the difference is that the super now has somewhere clean to live instead of landing across the subject. This closes the loop between Chapter 6 (composition), Chapter 19 (the interview), and this chapter: composition is graphics real estate, decided on set.

36 (Teach it back). A strong ~200-word explanation uses the five jobs, gives one concrete informing graphic (a name super that tells you who's speaking and why to trust them) and one decorative one (a spinning logo bug that just fills space), and lands the test: turn the graphics track off — does the film still play? Grade on clarity for a true beginner and on whether the two examples genuinely illustrate "information" versus "noise."

38 (Recreate It). Grade on whether the student captured the approach rather than copied the content: the typeface's voice (serif/sans/display and what it signals), the placement (which third, safe margins), the palette (how restricted), and the easing (how it lands). A strong reflection names a specific thing that was harder to match than expected — usually the restraint ("I kept wanting to add a second color / a bigger move, and the original didn't"). The lesson: reverse-engineering the best exposes how much of professional work is what was left out, and trying to match it raises your own floor faster than any tutorial.

37 (The Project 3 graphics pass). This is the Production Checkpoint, so there's no single model answer — but a complete submission has all four parts: a filled-in style kit, an opening title (readable, title-safe, voice-matched, one eased move), a reusable lower-third template applied to every speaker (dead space, once, read-twice), and one motivated animated element (eased, gone when done). The self-check is the graphics-off / graphics-on watch: the film plays without the layer and looks finished with it.

Chapter 35 — Answers to Selected Exercises

Model solutions and critiques for the starred and odd-numbered items. Attempt each before reading.

A. Seeing the effect

1. Count the invisible. Look specifically for: skies that are too perfect for the light on the ground; screens (phones/TVs) that never flicker or glare; backgrounds behind talking heads that are suspiciously clean; wide shots where a modern element "should" be but isn't; reflections in windows/glasses with no crew. A polished commercial or drama will hide 5–20 invisible shots in ten minutes. The point of the exercise is the realization that VFX ≠ spectacle: most of it is removal, replacement, and cleanup you were never meant to notice.

2. Find the bad key. Common tells, in rough order of frequency: (1) green fringe/halo on hair and shoulders (despill not done); (2) hard "sticker" edge with no soft grey (matte over-clipped); (3) light mismatch — subject lit differently from the background they're keyed into; (4) chewed hair (matte eroded too far); (5) buzzing/chattering edge (uneven screen or bad codec); (6) no grade match — foreground and background are two different color worlds. Weather segments and low-budget ads are the richest source.

3. Where's the screen replacement? Convincing replacements have: correct tracking (no sliding/floating), glow spilling onto hands/faces, a soft reflection and a hint of glare, and a grade that matches the room. Unconvincing ones are razor-sharp, flat, matte, and throw no light — a decal on glass. The one-sentence difference: a real screen is a light source with reflections; a bad replacement is a sticker.

4. The light test. Ignore the subject and answer three questions about the background world: which direction is its main light, what color, and how hard/soft? Then check the foreground subject against those three. Where a composite fails, at least one won't match — most often the direction (subject lit from one side, world lit from the other). This is why §35.2 insists you light the subject for the plate.

5. Spot the synthetic. Map onto the three §35.6 questions: Consent — did the person plainly agree (an actor de-aged in their own film: yes; a public figure faked in a clip: no)? Impersonation — could a viewer believe it's genuinely them saying/doing this? Disclosure — was it labeled/credited, or presented as real? A strong answer identifies a legitimate, disclosed, consented use (e.g., a credited de-aging or a licensed voice) versus a non-consensual, undisclosed, deceptive one, and explains which questions each passes or fails.

B. Reading composites

6. Name the layers. Foreground = the presenter (RGB). Matte = white over the presenter's body/face, black over everything else, thin grey at the hair/edges. Background = the weather map. In the composite, the black lets the map show through and the white keeps the presenter. Note that the presenter is usually shot on green and keyed — the "map behind them" is the plate.

7. Diagnose the field. Rewritten THE LIGHT: "Hard key from camera-right on the subject; the plate stays soft and warm from the left — two incompatible lights sharing one frame." Rewritten THE EFFECT: "The viewer can't say why, but the person looks pasted onto the room rather than standing in it — the eye reads the light mismatch as 'fake' before it reads any edge." Lesson: matching light is more important than a perfect matte; the eye reads light first.

8. Read the matte. The one problem: the background is dark grey with texture, not pure black — meaning some green will show through as a haze in the composite. On-set cause: the screen wasn't lit evenly (dark corners/wrinkles gave a second, darker green value the keyer under-selects). Post fix: clip the black (push near-black to fully transparent) and/or use a garbage matte; on set next time, light the screen flat. The white subject and clean grey hair are fine — don't touch them.

9. Write your own. A strong answer picks a genuine invisible composite (a set extension or screen replacement), fills all seven Described Shot fields, and in the eighth note names what was composited and why it's undetectable (matched light, hidden seam behind motion/focus, correct glow on a screen). If you can't tell how it was done, say so — that's the effect succeeding.

C. Shooting the key

10. Hang a screen. Guidance: judge evenness by eye and, if you can, on a waveform (Ch.5) — the green should read as close to one flat value corner to corner. Feathering lights (aiming their softer edge across the center) evens a hot middle; pulling lights back and adding distance evens hotspots. Kill wrinkles first — they're the cheapest fix and the biggest offenders.

11. Two lights, two jobs. Model self-critique: "Screen reads even; subject's key matches my intended warm room. But I see green spill on the left shoulder and a green kick in the hair — I was 4 ft from the screen, too close. Fix: move to 7–8 ft and add a rim light. The face and fill are good; the edges are the problem, and edges are a separation/spill issue, not a key-strength issue."

12. The distance experiment. Expected result: the close clip keys with green spill on the edges (hard to remove without eating the subject) and often a shadow on the screen behind the head that leaves a grey patch; the far clip keys cleaner with little spill and no shadow. The lesson lands physically: no post control fixes a subject who was standing in a green room. Distance is free and decisive.

13. Shoot for the plate. Guidance: write down the plate's light direction, color, and hardness first, then match your subject key to it. Shooting to a known target removes the hardest part of compositing — guessing what will match. If your plate's sun is camera-left and hard, your subject key is camera-left and hard. Same for color temperature and contrast ratio (Ch.11).

14. The clean plate habit. Guidance: the clean plate must be the same locked camera, same light — it's only a free patch if nothing moved. In post, lay the clean plate under the shot and mask the object out; because the plate matches grain, light, and grade automatically (it is the shot), the removal is near-instant. This is "shoot for the edit" pointed at VFX.

D. Keying and compositing

15. Pull your first key. Model answer: after sampling green, switch to the matte view. You want pure black background, pure white subject, thin grey hair. Beginners are shocked how much is not pure — grey haze in the background (uneven screen), grey holes in the subject (reflective/green areas). Diagnosing in black-and-white is the whole habit; the color view flatters and hides these.

16. The five-step clean. Model answer: for a well-shot clip, despill usually makes the biggest visible difference (kills the green edge glow). For a poorly-lit clip, clip black matters most (removes background haze). The meta-lesson: the step that helps most tells you what the shoot got wrong — lots of despill needed = spill on set; lots of black-clipping = uneven screen.

17. Screen replacement. Model answer: track quality is everything — if the replacement slides or floats, the corner track failed (re-track, or use markers). Once it's locked, the selling is glow + reflection + glare. Self-test: cover the phone with your thumb on playback — does light still spill onto the hand where the screen was? It should. A replacement that throws no light reads as pasted.

18. Stock element, matched. Model answer: at full strength the light leak/flare screams "effect"; reduced to a hint and graded to the shot's warmth, it becomes atmosphere. The restrained version almost always wins because the goal is a feeling, not a visible overlay. Screen blend mode drops the black; opacity and a grade do the matching.

19. Roto by hand. Guidance: even 20–30 frames of roto is tedious — draw the shape, step forward, nudge every point to the subject's new position, repeat. The one-sentence takeaway writes itself: green screen exists to buy you out of exactly this labor. You roto only when you were handed footage that was never shot for a key.

E. Fixing the composite

20. Fix the fringe. Post fix: despill (neutralize the green on the edges). On-set prevention: more separation plus a rim light to overpower the spill with a clean edge. The exercise's point: the fringe is a spill problem, and spill is fought on set first, despilled in post second.

21. Fix the shoot. Diagnosis by stage — Pre: a wrinkled bedsheet and a tiny room were the wrong setup; plan a taut screen and enough space. Production: (1) one lamp = uneven screen → light flat with two soft sources; (2) subject against the screen = spill + shadow → move forward 6–10 ft; (3) no rim → add one to beat spill and separate; the dark hair being eaten is a spill-plus-tight-key problem. Post: clip black for the corner patches, despill the edges, garbage-matte the worst corners, soften gently — but note honestly that most of this is an on-set fix; the post can only mitigate.

22. Fix the sticker. Missing, and how to add each: tracking (if it floats, re-track the corners); glow onto the hand/face; reflection across the glass; glare highlight; slight motion blur if the camera moves; and a grade match so the screen's color belongs in the room. The principle: a real screen is a light source with reflections, not a flat image.

23. Fix the mismatch. Full fix: (1) flip/relight so the sky's light direction matches the foreground (sun on the same side); if the sky can't be flipped believably, choose a different sky that matches. (2) Grade the sky to the scene's color and contrast. (3) Defocus/soften and add grain to the crisp sky so it matches the grainy footage. General principle: every stock element must share the shot's light direction, grade, grain, sharpness, and motion blur — an unmatched element is worse than none.

F. Settings and scenarios

24. Read the box. The eight: screen (taut, even), screen light (two soft, 45°, equal), subject light (matched to plate + rim), exposure (clean even mid-green, not blown), shutter (1/50 s, 180° — motion blur match), lens/DoF (match the plate), white balance (set and lock), codec (least-compressed, highest chroma). Each reason in a phrase: even screen → one color to key; matched subject → belongs in plate; correct exposure → clean key; shutter → blur matches; lens → perspective/DoF match; WB → no drift; codec → clean color edges.

25. Write the capture settings. (a) Phone green-screen interview: flattest/highest-quality phone mode, 1/50 s at 24 fps, expose screen to even mid-green, subject 6–10 ft forward with a rim — the one thing: an even, wrinkle-free screen. (b) Product with screen replacement: normal exposure, a solid clean color card (not green if the product is greenish) where the screen goes, keep the shot fairly steady — the one thing: a clean, trackable screen and steady framing. (c) Fast subject to roto: highest shutter you can while keeping acceptable blur, good contrast between subject and background to ease masking — the one thing: separation of subject from background (tone/color) so the roto shape is easy to see.

26. Match the plate. Warm café by a window: soft key from the window's side, warm WB, low contrast (~1.5-stop fill), gentle. Cool modern office: harder, cooler key from above/side, higher contrast, maybe a cool rim. Overcast street: very soft, near-shadowless, flat top light, neutral-cool, minimal contrast — an overcast plate has almost no directional key, so don't add a hard one. Pull directly on Ch.11–13.

G. Recreate It & Five Ways

27. Recreate the invisible. Guidance: the split-screen "one person twice" (locked camera, same light, mask the middle) is the most reproducible. Judge only by "can a friend tell?" If the seam shows, your light drifted between halves or the camera moved — the two failures the film's discipline prevents.

28. Five Ways: one composite, five matches. Trade-offs: the match that wins is usually the plate whose light you could most convincingly reproduce on your subject. A bright exterior is hardest (hard sun, high contrast — easy to mismatch); a soft interior is easiest (forgiving, soft light). What "sold it" is almost always the light direction and grade match, plus grain — not the key quality. The exercise proves matching beats keying.

H. The ethics of synthetic media

29. Run the three questions. (a) De-aging a consenting actor in fiction: consent yes, impersonation no (it's the actor themselves, in fiction the audience knows is fiction), disclosure optional → proceed. (b) Cloned politician "saying" something as news: consent no → STOP (and it's deceptive impersonation and disinformation — wrong on every axis). (c) Consented synthetic voice restoring narration: consent yes, not deceptive (it's their restored voice), disclose if listeners might assume it's a live recording → proceed, likely disclose. (d) Real customer composited into a scenario they never filmed: consent for this specific use is the gate — without it, STOP; even with a general release, if it depicts them endorsing/doing something they didn't, it's deceptive → get specific consent and disclose the dramatization.

30. Write the disclosure. Example (documentary voice clone): "Some narration in this film uses an AI-generated version of [name]'s voice, created and approved by them, to complete lines they were unable to record." Placement: an on-screen note at first use and in the end credits. Why: it keeps the audience's trust by making the synthetic element knowable rather than hidden.

31. The consent record. Checklist to get in writing before shooting: a likeness/appearance release (their image in the composite and its uses/territories/duration); a voice release specifically covering synthesis/cloning and what new lines may be generated; explicit scope ("for this branded piece and its cuts," not "any use forever"); approval rights over synthesized lines; a disclosure agreement (how synthetic elements will be labeled); and records kept with the project. Tie to Ch.38 releases. The rule: if you can't get specific, documented consent for the specific synthetic use, you don't do it.

I. Interleaved & Synthesis

32. Interleave: light, capture, key. (1) Ch.11: light the subject to match the plate and add a rim to beat spill. (2) Ch.3: capture the least-compressed, highest-chroma format so color edges survive. (3) Ch.35: light the screen flat and even, with separation. Single rule: a clean key is an even screen + matched, rim-lit subject + a high-chroma codec — all decided on set.

33. Interleave: grade the composite. You return to Ch.32 because a composite is two images that must become one: the vectorscope/flesh-tone line keeps the keyed subject's skin honest; one shared grade unifies contrast and color temperature; matching grain unifies texture. A great grade rescues an okay composite by making both layers share a world; an ungraded composite always reads as fake because two un-unified color worlds sit in one frame and the eye catches the mismatch instantly.

34. The "should I?" audit. Guidance: this is the whole chapter as a decision. Any effect whose "why the story needs it" sentence is weak, vague, or really about showing off gets cut. What survives is a short, motivated list — usually one or two effects. That discipline is the difference between a finished piece and an over-decorated one.

J. Going further

37. The window replacement. Model answer: the tell of success is depth order — the new view must read as being behind the glass, not pasted on it. Sell that with: a soft glow/haze on the glass, the window frame's inner edge in front of the plate, a slight defocus on the view if the subject is sharp (a real distant view is often softer), and a grade matching the room's light coming through. If it floats, you've forgotten the glass — add the reflection/haze and the frame edge. This is a genuinely common paid job (dull or blown windows behind interviews), so it's worth getting right.

39. The invisible performance join. Model answer: this works only if the camera and light didn't move between takes — any drift makes the seam visible. Hide the seam in a still, low-detail part of the frame (a plain wall between two positions, or a vertical split where nothing crosses). Match the frame you cut on so lips/eyes are continuous. If someone can tell it's two takes, look for: a lighting shift (the takes weren't identical), a moving element crossing the seam, or a mismatched frame at the join. The lesson: the same mask-and-hidden-seam skill as a screen replacement, pointed at performance — exactly Case Study 1's FIGURE CS1.4.

35. Teach it back. Model (~200 words): "A visual effect isn't a spaceship — it's a fourth channel called alpha that stores, for every pixel, how see-through it is. View that channel in black-and-white and it's a 'matte': white means keep, black means drop, grey means a soft edge like hair. Compositing is just stacking pictures and letting each matte decide what shows. The 'best effect is one you never notice' because the second you see an effect, you stop believing the picture. And you make it invisible mostly on set, not in post: shoot your subject under the same light as the background you'll drop them into, keep the camera and lens matched, and capture the cleanest color you can so the edges key well. Then in post you hide the seam — despill the green off the edges, soften them a hair, and grade both layers into one world with matching grain. Example: in a famous drama, one actor plays identical twins, composited together in scene after scene — and nobody notices, because the two 'brothers' share one light, one lens, one grade. That invisibility is the craft."

Chapter 36

Chapter 36 — Answers to Selected Exercises

Model answers and critiques for the starred and odd-numbered exercises (and the even items flagged "Answer/Model provided"). Treat each as one strong response, not the only one. Every spec number is typical, at the time of writing — confirm against Appendix H. Numbering matches exercises.md.

1. Name the two halves. For each file: the extension is the container (.mp4, .mov, .mkv); the codec is inside and often not deducible from the extension. The files you can't judge from the extension prove Chapter 3's point: a .mov can hold ProRes or H.264; a .mp4 usually holds H.264/H.265 but not always. The box does not reliably name its contents — use a tool like MediaInfo to see the codec.

2. Watch the re-encode. Expect the platform's version to be slightly softer, with more banding in gradients and blocking in shadows, and loudness normalized (often quieter than your master if you mixed hot). The lesson: the platform re-encodes, and the result reflects the quality of the file you gave it — hence "upload generous."

3. Spot the caption type. Burned-in examples are typically vertical social clips (TikTok/Reels), where captions are always on and stylishly placed — chosen because the audience is sound-off and a toggle would go unused. Sidecar/closed examples are typically YouTube, streaming, and broadcast — chosen for toggling, multiple languages, search, and accessibility. Strong answers tie each choice to the destination's viewing context.

4. The loudness jolt. The too-loud maker most likely mastered hotter than the platform target (well above ~-14 LUFS) or ignored loudness entirely; the platform normalized it, but the relative jump against a correctly-mastered neighbor is jarring, and over-limiting to get loud can leave the audio crushed. Fix: master to the target (Ch.33), then QC the exported file's loudness.

5. Read FIGURE 36.1. The seven areas: (1) filename/location — findable later (Ch.37); (2) format/container — plays where it's going; (3) codec — right size/compatibility; (4) resolution/frame rate — MATCH the timeline or you soften/judder; (5) quality/bitrate — enough or you band/block; (6) audio — 48 kHz, healthy, right channels or the mix degrades; (7) range/captions — whole timeline, captions attached. Each careless setting maps to a specific, nameable failure.

6. Read the preset. To the skeptic: the platform re-encodes every upload, and a lossy encoder can only work from the quality you hand it — it cannot add back detail. Give it a clean, generous file and its compressed result still looks sharp; give it a starved, already-damaged file and it compresses your artifacts, stacking damage. You aren't wasting the extra data — you're buying the quality that survives the re-encode. A bigger upload is the price of a good-looking stream.

7. Decode a real spec. (Do-it.) A strong result is a filled FIGURE 36.2 with the platform's current recommended container, codec, resolution, bitrate, audio, and caption support for your resolution — with the lookup date noted. The date matters because the spec will change; the habit of dating a spec is the professional tell.

8. Two specs, one table. The vertical-social and broadcast rows should disagree most on: aspect (9:16 vs an exact broadcast frame), codec/container (H.264 .mp4 vs ProRes/.mxf), loudness (~-14 LUFS vs ~-23/-24), and captions (burned-in vs a required sidecar format). The "why different in kind": social is a forgiving, sound-off, phone, self-serve platform; broadcast is a strict, sound-on, engineered chain that rejects non-conforming files — one optimizes for reach, the other for technical conformance.

9. Your first deliberate export. Self-check passes if: the file plays in a plain player (not just your editor), looks like your timeline (no new softness/judder → resolution and frame rate matched), and runs full length with full audio. If it looks worse, the culprit is almost always a resolution/frame-rate mismatch or a starved bitrate.

10. One source, two specs. Any quality difference from the timeline is usually the delivery bitrate (banding/blocking = too low) or a resolution/frame-rate mismatch (softness/judder). On the phone, the vertical one must keep subject and captions inside the safe zone; the horizontal one should look as sharp as the timeline. The point: each home got its own build, not a reused preset.

11. Make and keep a master. Expect the ProRes/DNxHR master to be several to many times larger than a normal H.264 delivery (often an order of magnitude). That size is correct for a master (it is the near-lossless negative you print copies from) and wrong for a deliverable (viewers don't need it and it's slow to send). Keep the master; ship the small copies.

12. Derive, don't photocopy. The transcode-from-master and export-from-timeline versions should look nearly identical (the master is near-lossless, so one extra generation is almost invisible). The third — a deliverable made from a deliverable — looks worst: softer, more banded, more blocked. That damage is generation loss. The lesson made physical: always go back to the master or the timeline.

13. The full delivery package. A strong package has: (a) an archived master (ProRes/DNxHR or high-bitrate H.264), (b) a web .mp4 with a .srt, (c) a 9:16 with burned-in captions in the safe zone, (d) a one-line note per file naming its spec. Every deliverable QC'd on a phone and a big screen. This is a real client delivery in miniature — grade on completeness and on each file matching its intended home.

14. 4K down to 1080. Deriving 1080p directly from the 4K source (master or timeline) is cleaner than chaining exports (4K → then re-export that to 1080p), because chaining adds a lossy generation before the downscale. Downscaling from a high-res, high-quality source also tends to sharpen perceived detail; downscaling from an already-compressed file carries its artifacts down with it. Derive from the highest-quality source, once.

15. The judder. Cause: frame-rate mismatch — the 24 fps timeline was exported at 30 fps, so frames were duplicated/interpolated and motion stutters. Fix: one field — set the export frame rate to 24 fps to match the timeline. (Nothing else needs to change.)

16. The eight-second film. They left an in/out range set (from reviewing an 8-second section) and the Range field exported only that marked section instead of the entire timeline. Fix: set Range to "Entire timeline." Nothing is corrupt.

17. The mush. Two compounding causes: (1) their export bitrate was starved, baking in softness/blocking before upload; (2) the platform then re-encoded that already-damaged file, compounding it. Fix: export at a generous bitrate (constant-quality "high" or the platform's number, 2-pass). The timeline looked sharp because it plays from full-quality media; the delivered file only has what the bitrate kept.

18. The silent scroll. Diagnosis: no captions on a sound-off audience — viewers meet a silent, contextless clip and leave; deaf/HoH viewers get nothing. Fix: add burned-in captions (not sidecar), because on autoplay-muted social a caption the viewer must enable is never enabled — burned-in guarantees they appear. Also confirm a strong sound-off hook in the first two seconds (Ch.23).

19. The rejected broadcast file. At least three likely-wrong areas: (1) loudness — -14 LUFS is streaming; broadcast wants ~-23 LUFS (EBU R128) or ~-24 LKFS (ATSC A/85) with a strict true-peak ceiling; (2) codec/container — broadcast usually wants ProRes (or a named codec) in .mov/.mxf, not H.264 .mp4; (3) resolution/frame rate — broadcast dictates an exact frame size and rate (e.g. 1080i/25 or 1080p/23.98), plus possibly bars/tone/slate and required sidecar captions. The correct process that prevents all of it: obtain and follow the broadcaster's delivery spec document line by line, and ask about anything ambiguous — never guess a broadcast spec.

20. The photocopy of a photocopy. Each new version was exported from the previous compressed deliverable rather than from the master or timeline, so each added a lossy generation. The phenomenon is generation loss. The fix: build and keep a near-lossless master, and derive every version (vertical, square, small) from the master or the timeline — each then one generation from pristine, none a copy of a copy.

21. The safe default. Container .mp4, codec H.264, a generous/constant-quality bitrate, AAC audio at 48 kHz. It's the safe universal answer because H.264-in-.mp4 plays on essentially every device, browser, and platform made in the last decade — "it will just play."

22. Four destinations. (a) YouTube talking-head: .mp4/H.264, 1080p, generous bitrate, sidecar .srt. (b) TikTok/Reels teaser: .mp4/H.264, 1080×1920 (9:16), safe zones, burned-in captions. (c) Wedding archival master: .mov/ProRes 422 HQ (or DNxHR HQX), full res/fps, clean picture, kept + backed up (Ch.37). (d) Client "website + maybe lobby TV": a universal 1080p (or 4K) .mp4/H.264 for the site — and keep the master so the lobby-TV/big-screen ask is served from a pristine source; ask them for specifics.

23. Bitrate judgment. No — not the same bitrate for equal quality. The confetti-filled, fast-moving event is far harder to compress (busy motion is the worst case, Ch.3), so it needs more data to avoid smearing/blocking; the locked-off interview against a plain wall compresses easily. This is exactly why constant-quality mode is the smart pick: it automatically spends more on the hard piece and less on the easy one, giving both the quality you want without you hand-tuning two bitrates.

24. The storage-squeezed master. Advise a high-bitrate H.264 master (as high a bitrate as storage allows, matched res/fps, clean picture). It's acceptable because a generous H.264 is far better to derive from than a starved delivery file and infinitely better than no master — it keeps future versions from being copies-of-copies. What they give up: it's not truly near-lossless, so it holds up to fewer re-grades/re-encodes than ProRes, and it's a weaker source for heavy future color work. Still: keep the best master you can afford.

25. H.264 vs H.265 decision. Case for H.265: ~half the file size for the same quality — meaningful for a big 4K travel film (faster upload, less storage). Case for H.264: universal playback, lighter to decode on any device, zero compatibility risk. Deciding factors: file size pressure, target-device decode support, and — crucially — that the platform re-encodes anyway, so for a platform upload the compatibility of H.264 often wins and the size saving of H.265 matters most for storage/transfer, not final quality. Recommendation: upload H.264 for maximum compatibility (or H.265 if the platform explicitly prefers it and size is a real constraint), and keep a ProRes master regardless.

26. The whole spec, from a brief. Files: (1) a master (ProRes .mov) archived; (2) a 90-second homepage hero — 16:9, 1080p (or 4K), H.264 .mp4, generous bitrate, burned-in captions because it autoplays muted (a sidecar nobody can toggle is useless there); (3) a 9:16 Instagram Reels cutdown — 1080×1920, safe zones, burned-in captions. The muted-autoplay detail changes captions to burned-in on the hero and argues the edit should carry meaning visually (no reliance on unheard VO for the hero), which may change the cut itself. Reading the brief changed both a delivery setting and an edit decision.

27. Write an SRT by hand. A correct .srt block is: a line number, a timecode line 00:00:01,000 --> 00:00:03,500 (comma before milliseconds), the caption text, then a blank line. Loaded against the video, the text should appear at the right times. It's universal because it's this simple — plain text, human-readable, editable anywhere, supported by nearly every player and platform. Complexity is the enemy of universality.

28. Auto then fix. Expect several errors even in a short clip: the proper name mis-transcribed, the number wrong or spelled out oddly, and words under the noise dropped or garbled. The three worst are usually the name, the number, and any phrase spoken over background sound. Proves the rule: auto-captions are a starting point; shipping them uncorrected gives deaf/HoH viewers an inaccurate transcript and reads as carelessness.

29. Burned-in vs sidecar, decided. (a) TikTok cooking clip → burned-in (sound-off, guaranteed visible). (b) Multi-language YouTube tutorial → sidecar (toggle + multiple language files). (c) Broadcast documentary → sidecar in the broadcaster's required caption format (spec-mandated, accessibility-compliant). (d) Archived master → sidecar kept alongside (master stays clean so captions can be corrected/translated/added later).

30. The full accessibility pass. A strong punch-list checks: captions present and hand-corrected; on-screen text legible on a phone and held long enough to read; no essential meaning carried by color alone (add label/shape/position); no risky flashing (≤3 flashes/second, or a warning); audio makes sense without seeing unspoken on-screen text (toward audio description). Fix at least the captions. The realization: accessibility is a short, concrete checklist done at delivery, not a vague ideal.

31. Recreate the master→deliverables flow. (Do-it.) A correct diagram has the master at the top with the actual deliverables the project needs beneath it, each labeled with its real spec (aspect, codec, captions), and arrows showing every deliverable derived from the master or timeline — none from another deliverable. Pinning it where you edit is the point: it makes the discipline automatic.

32. Interleave — capture meets delivery. The quality ceiling is set at capture: if the Café Scene was shot 8-bit 4:2:0 (Ch.3) and graded hard (Ch.32), the graded gradients are already near their banding limit on the timeline, and the delivery encode's second lossy compression pushes them over. If it was shot 10-bit, there's headroom for both the grade and the delivery encode. The delivery bitrate can preserve or waste that headroom, but it can never exceed the ceiling capture set. The chain is only as strong as its earliest link.

33. Interleave — the whole pipeline, one sentence each. Ch.2 (frame rate): the fps you shot and cut at is the fps you must export at, or the delivery judders. Ch.3 (bitrate/bit depth): what you captured sets the quality ceiling the delivery encode can't exceed and the banding it can worsen. Ch.23 (aspect/safe zones): the frame you composed for decides how it must be reframed and captioned for a vertical deliverable. Ch.31 (correction): the neutral, matched baseline you set is frozen into the master and every copy. Ch.33 (loudness): the -14 LUFS master you mixed is what the platform normalizes — deliver it faithfully or it gets crushed/quieted. The film's exit reveals every entrance.

34. Build your permanent checklist. (Do-it.) A good adaptation keeps the four phases (before export / export settings / after-export QC / before send), adds lines for your specific editor, gear, and platforms, and cuts irrelevant ones. Using it live and noting the first line that saves you is the exercise — that line is why professionals never deliver from memory.

35. Teach it back. A strong ~200-word explanation hits: the timeline is instructions, not a file, and export is a lossy translation of it; the codec/container distinction (what plays where); match resolution/frame rate or you soften/judder; the platform re-encodes so upload generous; and keep a master so versions don't become copies-of-copies (generation loss). One concrete failure per point (judder from a 30 fps export of a 24 fps cut; mush from a starved bitrate; a mangled Instagram cut made from the YouTube file). Teaching it cleanly proves ownership.

36. Deliver the portfolio. Self-audit rubric: each of the three projects has (1) an archived master, (2) a 16:9 web deliverable with a sidecar .srt, (3) at least one 9:16 social deliverable with burned-in captions inside the safe zone, (4) each file QC'd on a phone and a big screen for frame rate, loudness, captions, aspect, and glitches. Close any gap. When all three are complete sets of correct, captioned, spec-matched files, Project 3 is delivered and the delivery craft is done — on to Ch.37 to make sure the masters are never lost.

37. Five ways to deliver one clip. Expected trade-offs: (a) 16:9 web H.264 — the easy baseline. (b) 9:16 — you lose most of the width (reframe needed) and gain burned-in captions. (c) master — huge file, best quality, kept not shipped. (d) tiny email file — the hardest to keep looking good; an efficient codec (H.265) and a low target size force visible compression, so you accept softness for size and say so to the recipient. (e) 1:1 — a middle crop, loses less than 9:16 but still not the full frame. Hardest to keep good: the tiny email file (least data) and the 9:16 (most cropping). The lesson: every deliverable is a trade, and the master is the one you don't compromise.

38. Recreate a spec from a file you admire. (Do-it with MediaInfo.) A strong result reads out the real container/codec/resolution/fps/bitrate/audio and writes a matching preset. Common surprises: a great-looking file is often "just" H.264 in .mp4 at a generous bitrate (proving delivery is spec-matching, not exotic codecs); or a "master" someone shared is ProRes and enormous; or the frame rate isn't what you assumed. The point: you can reverse-engineer any delivery by reading the file, and most good deliveries are unglamorous, well-chosen defaults.

39. Recreate a platform's look, then beat it. Conclusion to reach: the platform applies the same re-encode to both uploads, but the higher-bitrate, cleaner source survives it visibly better — less banding and softness — because the re-encode can only work from the quality you give it. This is the §36.2 "upload generous / why it works" principle proven with your own eyes, and the single most convincing argument against exporting stingy "to help the upload."

40. The delivery capstone. A strong delivery set: an archived master (ProRes/DNxHR or high-bitrate H.264); a 16:9 web .mp4 with sidecar .srt; a 9:16 social .mp4 with burned-in captions in the safe zone; and a one-page delivery sheet listing each file, its exact spec (container/codec/res/fps/bitrate/loudness), its caption treatment, and its intended home — every file QC'd on a phone and a big screen. Grade on: does each file match its destination, is every deliverable one generation from the master, is everything captioned, and could a stranger read the delivery sheet and know exactly what each file is for? This is a professional client handoff; treat it as the rehearsal for the Production Checkpoint.

Chapter 37 — Answers to Selected Exercises

Model solutions and critiques for the starred (⭐⭐⭐) and odd-numbered exercises, plus the items flagged "(Answer provided.)". Yours will differ in specifics; judge against the reasoning, not the wording.

A. Auditing your own chaos

1. The one-copy census. The exercise's job is to make the danger concrete. A correct outcome is a list where at least one project is circled (one copy, one place) — almost everyone has some. What to do first: take the circled project you would least like to lose and give it a second, off-site copy today (a cloud upload or a drive taken elsewhere). Do not try to fix everything at once; the single highest-value action is converting your most precious "one copy" into "three copies, one off-site." Everything else in the chapter is refinement on top of that first move.

3. Name the failure mode. (a) Stolen backpack with laptop + its backup drive → violates 1 off-site; both copies were in one location. (b) Synced cloud folder where a deletion vanished everywhere → sync is not a backup (it propagates deletions; keeps sameness, not history). (c) RAID that survived a dead disk then was hit by ransomware → RAID is not a backup (it gives uptime, not a separate copy; ransomware hit the whole array). (d) "Had the backup" but never verified → the copy was never a verified copy (Ch.27); it may have been silently corrupt or incomplete. Each maps to one defense: off-site, real backup with history, an independent copy elsewhere, and verification.

B. Building the structure

5. Build the template. A correct template is the empty tree from FIGURE 37.1: numbered 01_FOOTAGE … 08_DOCS, a blank 00_README.txt, and (optionally) empty FONTS/ and LUTS/ inside 04_GRAPHICS. The grade is not the specific folders — any sensible set works — but that it is empty, saved, and reusable, and that you commit to duplicating it for every future job. If you found yourself "designing" it fresh, you missed the point: the value is freezing one and never redesigning.

7. Scale it down. For a 30-second phone talking-head: 01_FOOTAGE holds the clips with no camera sub-folders; 02_AUDIO is empty (single-system phone audio lives with the clip) or holds one room-tone file; 03_MUSIC-SFX holds one track + its license; 04_GRAPHICS maybe empty; 05_PROJECT the project file; 06_PROXIES likely empty; 07_EXPORTS one deliverable + a master; 08_DOCS a release if a person is on camera. The proof: the same skeleton works — small jobs leave branches empty rather than inventing a new structure. You never decide whether a project "deserves" structure.

9. The library level (⭐⭐⭐). A strong answer names a grouping and justifies it. Common good choice: group by year, then name each project DATE_CLIENT_what — because year-then-date sorts chronologically and dates group a client's jobs naturally; add a top-level _CLIENTS/ index only if you have repeat clients you think of by name. The _TEMPLATE_PROJECT lives at the library root (easy to find and duplicate); _ARCHIVE/ also at root (wrapped jobs move here off the fast drives). Justifications should cite sorting (dates/ISO), findability (one obvious home), and never the Desktop. Weak answers group by vague categories ("stuff," "clients," "misc") that overlap, so a project could live in two places — the test of a good scheme is that every project has exactly one correct home.

C. Names that survive

10. Fix these names. (a) Final video FINAL (2).mp4 → PROJECT_web-1080p_v02.mp4 — kill "Final" (×2), the space, and the parenthetical; use _vNN. (b) clip 7.mov → 2026-05-14_PROJECT_broll_subject_t07.mov — no space, zero-pad, add date/context. (c) Interview 4-3-26.wav → 2026-03-04_PROJECT_rec_int_t01.wav — ISO date (and note 4-3-26 is ambiguous: April 3 or March 4?). (d) logo copy copy.png → PROJECT_logo_v02.png — no spaces, version instead of "copy copy". The through-line: no spaces, no "final/copy," ISO dates, zero-padded numbers, self-describing.

11. Sort test. A computer sorts take1, take10, take11, take2, take3 — because it compares character by character, so 1 then 0 (take10) comes before take2. Fix: zero-pad → take01, take02, take03, take10, take11, which sorts correctly. The bug in one sentence: unpadded numbers sort lexically, not numerically.

12. Design your convention. A correct answer is a filled-in copy of the §37.2 Settings Box using a real project code — e.g. for "AURORA": 2026-05-14_AURORA_int-A_founder_t03.mov, 2026-05-14_AURORA_rec_int_t03.wav, AURORA_edit_v07.drp, AURORA_youtube-1080p_v03.mp4, AURORA_MASTER_prores-1080p_2026-05-28.mov. The grade: consistency and that you freeze it. Any coherent scheme beats a brilliant one you don't reuse.

13. Version discipline drill. Correct: edit_v01 (first cut) → edit_v02 (revision) → edit_v03 (second revision) → edit_v04 (client notes) → edit_v05_LOCK (locked). The wrong history: edit.drp → edit_final.drp → edit_final_v2.drp → edit_FINAL_clientnotes.drp → edit_FINAL_final_USE-THIS.drp — where nobody can tell which is newest or which was sent. The point: two climbing digits carry the entire history unambiguously; "final" destroys it.

14. Name a whole shoot (⭐⭐⭐). Success = 15–20 clips renamed so that, sorted by name, they group by setup and read in order (e.g. all int-A together, all broll_* together, takes ascending). Do it inside the editor or before import — never by renaming imported files on disk (that causes media offline). The real deliverable is the realization in the reflection: renaming took a few minutes and the next edit is meaningfully faster because you never hunt for "the good take" — you read it off the sorted list.

D. Never losing it (3-2-1)

15. Count your copies. A pass is three filled lines: three copies, two media types, one off-site — named specifically ("working SSD," "backup HDD," "cloud"). If any line is blank, that blank is the exercise: set up the missing one now (usually the off-site). A subtle fail is three copies that are secretly one location (all in the same room/bag) — that is one copy for disaster purposes.

17. Build your 3-2-1. Judge on concreteness and realism, not ambition. Budget tiers: floor — phone + one cheap external + free-tier cloud (already 3-2-1). Solo — working SSD + backup HDD + paid cloud or an HDD kept off-site. Volume — NAS for working+local backup + cloud/LTO off-site. The mark of a good answer: the person can and did set up at least the off-site copy today, because a plan not executed protects nothing.

18. Two-drives-in-a-bag autopsy (Answer). Still wrong: both copies share one location, so a stolen bag, a spill, a surge, or ransomware reaching both drives via the same laptop takes everything at once — that is 3 copies of the vulnerability, not 3 independent copies. It also may be only 2 copies (below the "3"), and the bag copy may be a sync (propagating deletions) rather than a true backup. The one change that fixes most of it: add an off-site copy (cloud, or a drive kept elsewhere) — the "1" is precisely the piece that survives a whole-room disaster.

19. Sync trap (Answer). Sync's job is sameness — it makes every location identical, so when you deleted the folder it faithfully deleted it everywhere, instantly. A real backup keeps history: older versions and deleted files you can roll back to. Tell the beginner: "Sync mirrors your mistakes; backup remembers what things were before the mistake." Use a versioned/backup service (or periodic separate copies you don't overwrite) for the off-site "1", not a bare sync folder.

20. Disaster tabletop (⭐⭐⭐). A strong answer traces each disaster against the actual setup. Dead SSD → survive if a backup + off-site exist; lose nothing but time. House fire → only the off-site copy survives; if there is none, total loss — this is the disaster that most often exposes people who "had a backup" locally. Ransomware → anything writable and connected can be encrypted, including a plugged-in backup drive and even a live-sync cloud folder; the survivor is an offline/disconnected copy or a versioned backup you can roll back. Where any disaster wins, the fix is almost always "add or disconnect an off-site/versioned copy." The learning: 3-2-1 isn't one rule, it's a defense sized to the worst plausible event, which is the whole-room one.

E. Drives and speed

21. Match drive to job (Answer). (a) active edit → external (or internal) SSD (fast playback). (b) scratch/cache → internal NVMe SSD, separate from the only media copy. (c) local backup at rest → external HDD (cheap, speed irrelevant). (d) off-site → cloud (or an HDD kept elsewhere). (e) deep archive → HDD (×2) or LTO tape (cheapest per TB, store-and-verify). Principle: fast/expensive for what you work on, cheap/slow for copies at rest.

22. Diagnose the stutter (Answer). In order: (1) Drive — is the active media on a fast SSD, not a slow HDD or a full system drive? Move it. (2) Port/cable — is the SSD on a fast port? A modern SSD on an old USB port crawls; two identical-looking ports can differ tenfold. (3) Codec — is it heavy Long-GOP (H.265) that's hard to decode (Ch.3)? Then generate proxies (Ch.27) — the file-side fix. (4) Cache location — is the scratch/cache on a fast drive with free space, not choking a full system drive? The drive-side and file-side fixes (fast disk + proxies) together solve nearly every stutter.

23. Place the scratch (Answer). Put the scratch/cache on a fast drive that is not the drive holding your only media copy — ideally a fast internal or a dedicated fast external — both to spread the read/write load and so a filling cache can't choke the drive your media lives on. It's safe to simply delete it because the cache (render files, thumbnails, waveforms, proxies) is entirely regenerable — the editor rebuilds it on demand. Never back it up; never let it fill your system drive.

24. Spec a rig (⭐⭐⭐). A defensible solo rig: working = a 1–2 TB external SSD on the fastest port (edits smoothly); scratch/cache = the internal NVMe (fast, disposable), kept clear of the media copy; local backup = a larger (4–8 TB) external HDD (cheap, at rest); off-site = paid cloud or a second HDD kept elsewhere; archive = wrapped jobs onto HDD(s) in _ARCHIVE, two copies, one off-site. Every choice must be justified by speed (SSD for the edit and cache), cost (HDD for at-rest copies), or safety (a second medium + off-site). Capacities are illustrative; the roles and reasoning are the grade.

F. Archiving

25. Keep or drop? (Answer). Keep: (a) camera originals — irreplaceable; (d) master export — the thing you re-master from; (f) fonts — or titles reflow in a year; (g) music license — proves your rights (Ch.38). Drop: (b) proxies — regenerable from originals; (c) render cache — the editor rebuilds it; (e) the seven draft exports — keep only the master + final deliverables. The rule behind every call: keep the irreplaceable, drop the regenerable.

26. Write the manifest. A model 00_README.txt: project name and client; cut in <editor> v<version>; frame rate and resolution (e.g., 1080p, 25 fps); location of the master (07_EXPORTS/PROJECT_MASTER_...); fonts used; LUTs used (or "none"); music track + where its license is; delivery specs and date delivered; any account/password dependency; your contact. The test: a stranger (future-you) could reopen and finish the project using only this page.

27. Archive vs backup (Answer). Model (~120 words): "A backup protects work I'm actively doing, right now — it's a live copy I keep updating, so if my drive dies mid-edit I lose nothing. Example: my current project synced to a backup drive and the cloud every day. An archive preserves work I've finished — I make it once, at wrap, strip the disposable files, and store it long-term so I can reopen it in years. Example: last month's delivered film, consolidated with its originals, master, and licenses into one package on two drives plus the cloud. Backup is for the doing; archive is for the done. A backup is a net under the trapeze; an archive is the film in the vault."

28. Archive one project for real (⭐⭐⭐). The whole point is the reopen test: after consolidating, stripping, writing the README, and 3-2-1'ing, you rename/unplug the working copy and open the project from the archive alone. A correct report names whatever came up "media offline" and how it was fixed (usually: a file that lived outside the project folder and wasn't consolidated — add it and re-verify). If nothing came up offline and titles/grade are intact, the archive passed and is trustworthy. An untested archive is a hope, not an archive.

G. Handoff and portability

29. Why offline? (Answer). A project file stores only pointers to where its media lives, not the media itself; emailed alone, those pointers reference files the new machine doesn't have, so everything is "media offline." General fix: send a consolidated, self-contained package (the editor's project-archive/collect function) containing the project and its media, referenced by relative paths.

30. Spot the missing dependencies (Answer). Wrong title font → the fonts weren't included, so the new machine substituted. Flat grade → the LUT the grade relies on is missing. Effect "offline" → a plugin/effect the other machine doesn't have installed. The package should have included the fonts and LUTs (in 04_GRAPHICS), a note listing required plugins, plus all media and a README — everything the video secretly depends on, not just the visible clips.

31. Package it (Answer). Success = the consolidated package opens in a different location/machine with media, fonts, and grade intact. Common gaps to expect: media that lived outside the project folder and wasn't collected; fonts and LUTs (frequently forgotten); plugins that must be installed separately; linked graphics/music source files. Note whatever broke — that's your personal checklist of what to include next time.

32. Relative vs absolute (⭐⭐⭐). Model analogy: an absolute path is like telling someone your friend lives "at 12 Oak Street, Springfield" — true only in that one town; move the whole neighborhood and the address is wrong. A relative path is "in the house next to mine" — still true wherever the block is picked up and set down, because the two houses moved together. Keep media inside the project folder and reference it relatively, and project + media are neighbors that travel as one — so a copied folder relinks automatically. The one habit that makes portability automatic: media lives inside the project folder from day one (FIGURE 37.1), never scattered across drives.

H. Interleaved & synthesis

33. From card to archive (Answer). Order and the failure each step prevents: offload (verified, two copies) — a wiped/corrupt card losing the shoot (Ch.27); back up to 3-2-1 — a single failure losing everything (§37.3); ingest/import into the structure — media the project can't find (Ch.27); organize folders→bins, named — "can't cut what you can't find" (§37.1–37.2, Ch.27); sync double-system audio — tedious blind syncing later (Ch.27); proxy if heavy — a stuttering edit (Ch.27); selects → string-out → cut — no shape (Ch.27–29); deliver master + deliverables (Ch.36); archive consolidated + 3-2-1 + README — an un-reopenable project a year later (§37.5). Each step prevents a specific, nameable loss.

34. The master's journey (Answer). The Ch.36 master is a high-quality, lightly-compressed version you re-master everything else from; it's what you archive because from it you can make any future deliverable (a new platform spec, a re-edit, a reel clip) at full quality. Archiving only the delivered YouTube file leaves you with a heavily-compressed dead end — you can play it but not re-master or re-edit from it well, and you'd have no originals to change the cut. In Ch.39 you'll pull clips from these archives for your reel; that only looks good if you kept the master (and the originals), not the compressed deliverable.

35. The paperwork (Answer). Into 08_DOCS: the brief, the signed model/property releases (Ch.38), the invoice, and delivery notes. Into 03_MUSIC-SFX: the music/SFX license files. A technically perfect archive without these is still unusable legally: you can't safely re-publish or repurpose the piece if you can't prove you had the right to the faces and the music in it. Rights are part of the asset; an archive that preserves the pixels but not the permissions preserves a liability.

36. Teach it back (⭐⭐⭐). A strong ~200-word answer makes the core claim — never losing work is a habit, not a talent — and supports it with three concepts and a concrete disaster. Model gist: "Losing work feels like bad luck, but it's actually a missing habit. Habit one, structure: every project goes in the same folder tree, cloned from a template, so nothing is ever 'somewhere on my desktop.' Habit two, 3-2-1: three copies, two media, one off-site — so when a drive dies (and drives die on a schedule), it's a shrug, not a catastrophe; the editor who lost a wedding had one copy on one card. Habit three, archive: at wrap I consolidate the originals, project, and master into one package, strip the junk, and store it in three places with a README — so a year later I can reopen and change it in an hour. None of this is talent. A brilliant editor with one copy of their footage is one click from losing everything; an average editor with these habits never does. The difference isn't skill at cutting — it's boring habits, done every time, that make loss almost impossible."

37. The Production Checkpoint, committed (⭐⭐⭐). No fixed answer — the deliverable is three archived, 3-2-1'd, reopen-tested projects and, for each, a confident one-sentence recovery plan ("if my main drive died tonight, I'd pull the archive from [off-site] and re-copy it to a new drive tomorrow"). If any recovery sentence is uncertain, that project isn't safe yet — fix the gap. Confidence across all three is the proof you've crossed the line this chapter is about.

Chapter 38 — Answers to Selected Exercises

Model solutions and critiques for the starred (⭐⭐⭐) and odd-numbered exercises. Yours will differ in specifics — especially the dollar figures, which depend entirely on your market — so judge against the reasoning, not the wording. Standing reminder: none of this is legal advice, and laws vary by jurisdiction.

A. Seeing the business

1. Spot the license. A correct outcome is noticing the pattern: professional and brand videos almost always have visible licensing (a track credited in the description, a library name, "music licensed via…"), while amateur uploads using popular songs typically have nothing — which is the tell that the music may be unlicensed and living on borrowed time (a claim or mute away from trouble). The lesson: visible licensing is a marker of professionalism, and its absence on commercial-looking work is a red flag, not a neutral omission. If you can't tell where the music came from, neither can the platform's dispute system when it flags you — which is why you keep the proof.

3. Reverse-engineer a price. The graded skill is the method, not a number. A strong answer breaks the video into three stages and estimates days: e.g., a small-business testimonial ≈ half a day pre-pro (planning, coordinating), half a day shoot, and 1–1.5 days post (edit, color, sound, captions, plus a vertical cut) — roughly 2–2.5 working days, before hard costs (a music license) and margin. The key insight to demonstrate: the edit is usually the largest single chunk, and the visible shoot is a minority of the work. Whatever your market's day rate, the price is that rate × the honest day estimate + costs + margin — not "it's a short video, so it's cheap."

B. Pricing

5. Iceberg audit. At least eight costs an employee does not pay from their wage but a freelancer pays from their rate: (1) gear purchase and depreciation/replacement; (2) software subscriptions; (3) business/liability and equipment insurance; (4) self-employment taxes and set-asides; (5) the unbilled admin — quoting, invoicing, email, scheduling; (6) marketing and portfolio time; (7) the unbooked days between jobs (you're paid per booked day, not per week); (8) sick days, holidays, and downtime with no paid leave; (9) retirement/benefits an employer would part-fund; (10) pre-production and revision time not always separately billed. The point: an employer subsidizes all of this, so a wage is take-home; a freelance rate must contain it, which is why matching an hourly wage is a slow bankruptcy.

7. Three models, one job (café testimonial). Day rate: quote your day rate × the estimated days (≈2–2.5) + the music license — you're pricing your time. Project rate: fold that internal estimate into one flat number the café can budget, with an INCLUDED/NOT INCLUDED block so the flat price has a fence. Value-based: invent a plausible goal — "this testimonial anchors the homepage and a paid ad campaign expected to drive online bean sales through the holidays" — and price against that outcome, which justifies a higher fee than "2.5 days," because the video is an asset earning for months. The exercise succeeds if you show the same job legitimately carrying three different numbers depending on the lens, and can say when you'd use each (day rate when crewing/unknown length; project rate for most defined client jobs; value when you understand and can price the outcome).

9. Price your three projects (⭐⭐⭐). A strong submission gives, for each of Project 1 (talking-head), Project 2 (doc short), Project 3 (branded piece): an honest day estimate across all three stages, the hard costs (music, any stock), a project-rate number built from your market's day rate, and one sentence naming the model and what it covers. Expect the numbers to rise across the three — Project 3 (motion graphics, grade, designed sound, more deliverables) is several times the work of Project 1. The most common error to catch in your own answer: pricing only the shoot and forgetting that Projects 2 and 3 are edit-dominant. This is part of the Production Checkpoint; keep the numbers, and the one-sentence justifications, on file.

10. The raise (⭐⭐⭐). A sound case has four parts: (1) what changed — your skill, speed, portfolio, and demand (being fully booked is the market telling you your price is too low); (2) your costs rose — gear, software, insurance, cost of living; (3) the new number and how you got there — a specific percentage or figure tied to the above, benchmarked against your market, not guessed; (4) how you tell existing clients — with notice, framed around continued value, applied to new projects (grandfathering a current job is goodwill). The deeper lesson: being fully booked at your current rate is not success, it's evidence of underpricing — the correct response to "I have more work than I can handle" is to raise prices, which either increases income or frees time, both of which are wins.

C. Contracts and the SOW

11. Clause recall. Eight-plus clauses and the question each answers: Parties (who's agreeing?); Scope/deliverables — the SOW (exactly what will you deliver, and what's excluded?); Price & payment schedule (how much, and when — deposit/milestone/final?); Timeline (shoot and delivery dates; what's owed when?); Revisions (how many rounds included?); Ownership/rights (who owns footage and project files?); Usage/license grant (where and how long may the client use it?); Releases & asset licensing (who's responsible for getting/clearing them?); Cancellation/kill fee (what's owed if the client pulls out after booking?); Late payment (fees; when do final files release?). Full credit = eight clauses each paired with the real-world disaster it prevents.

13. Critique the SOW. The described SOW — "Make some social media videos for our launch. A few revisions included. Delivered when ready." — is a scope-creep machine. Ambiguities and rewrites: "some social media videos" → specify the count, length, and aspect ratio ("3× videos, each under 30s, 9:16"); "for our launch" → name platforms and any must-includes/CTA; "a few revisions" → "two rounds, each a consolidated set of notes"; "delivered when ready" → fixed dates ("first cut by [date], final on approval"). Every vague phrase is a future argument; a specific SOW is a fence. The lesson: vagueness in an SOW is not flexibility, it's an unbounded liability you volunteered for.

15. Payment schedule design (⭐⭐⭐). For a larger, three-day, two-month branded piece, a defensible schedule: 50% deposit to book (it reserves three shoot days and funds pre-pro; often non-refundable because you're turning down other work), 25% at a milestone (e.g., picture lock or first full cut — it de-risks the long post stretch so you're not carrying two months of labor unpaid), and 25% on delivery (released with the master on final payment). Justifications should cite: the deposit protects the front (cancellation, held dates), the milestone protects the middle (long unpaid post), and delivery-on-payment protects the back (leverage). A schedule that's all-at-the-end on a two-month job means financing your client for two months and hoping — not a business, a loan.

16. Read the deal for traps (⭐⭐⭐). The four missing clauses and the disaster each prevents: (1) Revisions — without a cap, "revisions until happy" can run for weeks unpaid; add "two rounds, then change orders." (2) Ownership of files — without it, an awkward post-delivery fight over "send us the raw files"; state that you retain raw/project files (or price a buyout). (3) Music/asset licensing responsibility — without it, ambiguity over who's liable when a track gets claimed; state you clear what you bring, the client warrants what they supply. (4) Cancellation — without it, you shoot nothing and earn nothing if they pull out after you've held the dates; add a kill/cancellation fee. Each is one sentence; each prevents a specific, common, expensive mess.

D. Scope and revisions

17. Revision or new request? (a) wrong shade of blue on the logo → revision (a fix to the agreed deliverable). (b) "also get a square version" → new request (a new deliverable/aspect ratio). (c) "tighten the middle" → revision (refining the agreed cut). (d) "different music and a new voiceover" → new request (a substantial re-edit / new elements). (e) misspelled founder's name → revision (a fix — and one you should catch yourself). The discriminator: does it refine the thing you were hired to make (revision, included) or add/replace a thing (new request, billed)?

19. Draft a change order. A correct model:

Change Order #1 — governed by the Agreement dated [date]. New work: two (2) additional 15-second cut-downs (formats: 1× 1:1, 1× 9:16) for [platforms], not included in the original SOW. Added cost: [your rate for ~0.5–1 day]. Timeline: +[N] business days from written approval. Proceed on your approval below. [signature/approval line]

The grade: it names the specific new work, attaches a price, states the timeline impact, and requires approval before proceeding. It reads as friendly and matter-of-fact, not punitive — a yes-with-a-price.

21. Five asks, five responses (⭐⭐⭐). A strong answer invents five plausible out-of-scope asks and, for each, decides absorb-as-goodwill vs. change-order, with reasoning. Reasonable pattern: absorb the tiny, one-off, relationship-building things (a 10-second trim, fixing a typo you arguably should have caught) as cheap goodwill; change-order anything that is a new deliverable, a new version, a re-shoot, or repeated (extra platform cuts, a different song, a longer edit). There's no single correct split — the graded skill is a consistent, defensible policy rather than case-by-case resentment, and the recognition that "goodwill" is a deliberate, occasional investment you choose, not a default you get talked into. If you're absorbing more than a token amount, you've mislabeled new requests as revisions.

22. The never-satisfied client (⭐⭐⭐). These are genuine revisions (not new requests), but past the included rounds — so the tool is the lock, not the change order. What to say: warmly, "We've done our two included rounds and I want to get this locked so it can go live — let's do one final consolidated pass, and I'll consider that our lock; further changes after that we can handle as a small paid revision round." What the contract should have said: a defined number of rounds and an explicit lock point ("after round two, the cut is approved; further changes are billed"). How you get to lock: consolidate the remaining notes into one final pass, deliver it, and declare the lock. The lesson: "revisions until satisfied" has no floor; a job that cannot end cannot be profitable, and the fix is a boundary set in the contract, kindly enforced now.

E. Releases

23. Consent to film vs publish. Consent to film = the person agrees to be recorded in the moment (they stood still, they nodded). Consent to publish = the person grants documented permission to use that footage in specific ways — in an ad, on a company's channels, indefinitely. Example of film-only: someone at a counter says "sure, go ahead" as you record B-roll. Example of publish: that same person signs a model release granting use of their likeness in the finished commercial. The first does not include the second; commercial use of an identifiable person generally needs the second, in writing.

25. Release triage. (a) paid actor in a commercial → yes, model release (they sign). (b) wide crowd at a public festival → generally notice (signage/announcement) rather than individual signatures, since nobody is featured. (c) one featured face pulled from that crowd for a promo → yes, model release from that person (featured + identifiable + commercial). (d) 12-year-old in a school video → yes, model release signed by a parent/guardian. (e) recognizable interior of a private restaurant for an ad → yes, property/location release from the owner/manager. (f) stranger walking through deep background of a public-street shot → usually no individual release needed if incidental/not featured, though local law varies and "when in doubt, get it." The organizing test: featured + identifiable + commercial → release; incidental crowd → notice; minor → guardian; private property → property release.

27. The whole shoot's paperwork (⭐⭐⭐). A complete answer inventories Project 2 person-by-person and place-by-place and assigns the correct release to each: the interview subject → model release (they sign); any identifiable, featured people in B-roll → model releases; a minor anywhere → guardian signs; private interiors/locations → property releases; incidental background crowds → notice. Then the honest publish test: for every "no release yet," you cannot currently publish/sell/reel the piece around that person. The value of the exercise is the discovery of gaps now, while you can still collect signatures, rather than in the edit. Part of the Production Checkpoint — file everything in 08_DOCS.

F. Licensing and copyright

29. Cleared or not? (a) a song bought on a music store, under a client's ad → not cleared; buying a song for personal listening is not a license to use it in a video (and a hit needs sync + master rights anyway). (b) a track from a royalty-free library you subscribe to, in that same ad → cleared, if your subscription tier covers commercial/client use — check the tier. (c) a CC BY-NC track in a paid client video → not cleared; NC = non-commercial only, and client work is commercial. (d) a public-domain composition performed in a new recording you found online → not necessarily cleared; the old composition may be public domain, but the new recording can carry its own fresh copyright — you'd need that recording's permission. (e) a font "free for personal use" in a paid video's titles → not cleared; get a commercial font license or substitute an openly-licensed typeface. The through-line: match the license to the actual use, and remember credit/purchase/availability are not licenses.

31. The Content ID surprise. Steps: (1) don't panic — a claim is not a strike, and a false claim on licensed music is common; (2) locate your license for the track (saved in 03_MUSIC-SFX); (3) file the platform's dispute, submitting the license as proof of your right to use it; (4) keep the video's audio pending resolution. The one earlier thing that makes resolution possible: you kept the license record. Without the proof, you can't dispute; with it, a false claim is a formality. This is exactly why §38.5 insists on saving the license file — the day a system flags you is the day it earns its keep.

32. Rescue the unlicensed edit (⭐⭐⭐). Options, best to worst: (1) Best — swap to a properly licensed track from your library that fits the energy; yes, it changes the rhythm, so re-time the cuts (a real but bounded cost, and cheaper than a takedown). (2) License the actual song — usually impossible or wildly expensive for a hit (two rights, sync + master), so rarely viable for a small job. (3) Re-cut to a public-domain or CC-appropriate track if the piece can carry it. (4) Worst — publish unlicensed and hope — not an option; it hands the client a liability and risks mute/strike. Recommendation: option 1, done proactively before upload. How to tell the client without alarming them: frame it as a quality and safety upgrade ("I'm swapping in a properly licensed track so the video can never be muted or claimed on your channels — it'll sound just as good"), not a confession of error. The meta-lesson: this whole mess was preventable by licensing at the first drop; doing it right up front is cheaper than rescuing it.

G. Getting and keeping clients

33. Where work comes from. Ranked most→least effective: (1) referrals & repeat business from happy clients; (2) your network and a visible body of work (reel/portfolio); (3) being findable when someone searches; (4) cold outreach to strangers. Why the top beats any ad: a referred client arrives pre-trusted — someone with no incentive to lie has already vouched for you — so the sale is half-made and the price resistance is lower. No amount of paid advertising buys that trust, which is why the "boring" business discipline (fair pricing, clear contracts, defended scope, on-time delivery) is your marketing: it's what makes clients refer you.

35. Draft an invoice. A correct model includes: your name/business and contact; the client's name; an invoice number; the date and due date (net-15/30); itemized work (the package + any change orders); the subtotal, any tax, the deposit deducted, and the balance due; and how to pay (bank details/link) with a payment reference. A professional touch: "final master files delivered on receipt of payment." The grade is completeness and clarity — an invoice that looks like a real business's, not a number in a text message. Ambiguity here is how you end up unpaid.

36. Pitch a retainer (⭐⭐⭐). A strong pitch is framed around the client's benefit, not yours: "You've hired me three times this year — instead of quoting each one from scratch, here's a monthly retainer: [N videos / a package of social clips] each month for [flat monthly fee], with priority scheduling and a slightly better rate than one-offs. You get predictable content and a partner who already knows your brand; no re-briefing, no per-project negotiation." Why it's better for them: predictability, consistency (same person who knows their brand, Ch.22), priority, and usually a per-unit discount. Why it's the freelancer's dream (stated to yourself, not the client): recurring, predictable income that smooths feast-and-famine and deepens the relationship past the point of scope fights. The graded skill is selling the client's upside, not your cash-flow needs.

H. Interleaved

37. Brief to quote. A complete answer takes a Chapter 22 brief (goal, audience, key message, CTA) for a small business and produces (a) a one-line concept and (b) a quote with a NOT INCLUDED block — and shows the causal link: because you understand the goal is, say, "drive holiday online orders," you can price against that value (higher than "a day of filming") and scope precisely to it (a hero + a vertical cut aimed at the CTA), excluding everything else. The lesson: the brief (Ch.22) is what upgrades your pricing from time-based guessing to value-based confidence, and what makes the SOW specific enough to defend. No brief → no value pricing and a vague, creep-prone scope.

39. The archive as evidence (⭐⭐⭐). For a delivered project, 08_DOCS should hold every signed model and property release, the signed contract/SOW, and the invoice; 03_MUSIC-SFX should hold every music/SFX file with its license/receipt. Why a rights-clear archive is worth more than a rights-unknown one: a year later, when you want to re-cut the piece for a client's new campaign, repurpose a clip, or pull it for your reel (Ch.39), the archive proves you may legally do so — the releases show the people consented, the licenses show the music is cleared. A rights-unknown archive is a video you technically possess but cannot safely reuse, because you can't prove you have the right to; it's a liability with a timecode. This is the direct join between Chapter 37 (archive the paperwork) and Chapter 38 (the paperwork is the rights) — and why §37.5 put releases and licenses in the archive in the first place.

40. The whole business, one job (⭐⭐⭐). A model walkthrough hits every stage with the right tool: discovery — a call surfacing the Chapter 22 brief (goal/audience/message/CTA); quote — itemized across three stages with a music line item and INCLUDED/NOT INCLUDED (§38.1); contract + SOW + deposit — signed, with deliverables/revisions/ownership pinned and the deposit banked before work (§38.2); shoot — with model/property releases signed on the day and filed to 08_DOCS (§38.4); music — a royalty-free track licensed and logged in 03_MUSIC-SFX (§38.5); a revision — one out-of-scope ask handled with a warm change order rather than a free favor (§38.3); delivery — platform-correct exports with captions (Ch.36); invoice + follow-up — master released on final payment, then a warm note inviting a referral and repeat work (§38.6). Full credit = every stage named with its tool, showing the chapter as one integrated business process rather than six separate topics. This is the exam-in-miniature for the whole chapter.

Chapter 39 — Answers to Selected Exercises

Model answers and critiques for the starred and odd-numbered items. Grade on specificity and judgment, not on matching this text word for word.

1 (The ten-second test). Strong answers name the exact second and a specific cause: "I'd have left at 0:06 — five seconds of logo before any work," or "at 0:11 — the opening shot was a flat, generic drone shot that proved nothing." Good openers named: "a stunning shot on frame one," "immediate motion and clean light." The lesson: the decision is made in the opening, so the opening must be the best shot, not a warm-up.

2 (Weakest-shot hunt). The target skill is identifying the floor of a reel, not its ceiling. A strong answer names one specific shot ("the slightly soft handheld interview cutaway at 0:40") and reasons that cutting it would raise the whole reel's read, because impression floors at the weakest inclusion. If a student can't find a weakest shot, the reel is either excellent or they're watching as an audience, not an appraiser — push them to rank all shots and name the bottom one.

3 (Read the arc). Grade on whether they located the four movements and drew a curve, not a flat line. Strong answers note where cuts accelerate and where the peak sits relative to the music's crest. Common finding: weak reels have no arc (flat parade); strong ones visibly build. Some reels front-load and taper — valid, but note the reader should still see a deliberate shape.

5 (The niche read). The point is that a reel declares a niche whether or not the person intended it. Strong answers write a specific "hire them for ___" sentence per reel and notice mismatches ("the reel is all moody food but the bio says 'weddings and events'"). When words and work disagree, the work wins — a hirer believes what they see. This previews §39.5: let the reel state the niche.

6 (Picture the opener). The four things FIGURE 39.1 proves before a word: (1) framing/composition, (2) control of light, (3) intentional camera movement, (4) clean audio (via the one sync sound). Bonus if they add (5) taste — the shot was chosen to be first. The lesson: the opening is a dense proof, not a warm-up.

7 (Diagnose the field — the fast open). Opening on 0.5-second shots from frame one loses the anchor. THE CUT becomes "no held shot; immediate flurry"; THE EFFECT becomes "the viewer has nothing to lock onto — the reel feels frantic and no single shot lands as 'the best.'" What's lost: the slow→fast contrast that makes the acceleration feel like a build, and the chance for your single best shot to be seen. The hero shot needs air to do its job.

8 (Fix the timeline). Two problems: (1) the sign-off card in the middle interrupts the arc and hides contact where a mid-reel scroller won't return to it; put it last, held 4–5s. (2) Ending on a flurry cut to black feels like a crash, not a landing; resolve onto the card as the music lands. Fix: move the card to the end, let the music resolve under it, land the plane.

9 (Write your own reel-open). A full seven-field Described Shot for the student's actual best shot, framed as the opening 3 seconds. Grade on specificity and on whether they justify why it earns the watch (best shot + on-niche + held long enough to land). Weak answers describe a merely-decent shot or don't commit to it being first. The exercise forces them to name their hero.

10 (Run the funnel). Guidance: the win is that they threw away most of the pile at each pass and can state their survivor counts. If they kept 25+ at the end, they hoarded — send them back. A healthy result from three projects is roughly 34 → 19 → ~12–14 keepers (see CS-02). The discomfort of cutting good shots is the funnel working.

11 (Pick the hero). Judge the defense, not the pick. A strong answer commits to one shot and defends it in a sentence tied to impact and niche ("this plating shot is my most controlled light and instantly says 'food' — it earns the watch and states what I'm for"). "They're all good, I can't choose" is the failure mode; make them choose.

12 (Kill three darlings). Model: naming three shots they love that aren't among the genuine best, and cutting them with a reason ("the sunset timelapse is beautiful but off-niche and slows the build"; "the whip-pan is fun but shows off the move, not the subject"). The skill is separating "I'm attached to this" from "this earns its place." This is Ch.29's kill-your-darlings scaled to the whole body of work.

13 (Two reels, two lengths). Usually the 30s is stronger because the tighter constraint forced a higher floor. Strong answers report what the 30 exposed — shots that felt fine in the 90 but couldn't survive the 30 were probably always weak. The lesson: length is a curation tool; when in doubt, cut shorter and see what survives.

15 (Music first). Grade on whether they can articulate what each track does to the same footage and choose the one their work wants (not their personal favorite). The insight: music is a brand statement — a driving track and a warm piano make the same shots feel like different businesses. Choosing the track is choosing a positioning.

16 (Cut to the beat). The report should describe the felt difference: on-beat cuts feel "tight," "locked," "intentional"; off-beat cuts feel "sloppy," "loose," "amateur" — even with identical shots. The lesson (from Ch.29): rhythm is felt, not seen; cutting on the beat is disproportionately powerful and beginners skip it.

17 (Build the sign-off card). Model: name + niche (2–3 words) + one contact, legible, high-contrast, held 4–5s. The test question — "would a stranger know how to hire you from this frame?" — must be answerable yes. Failure modes: five social handles (splits attention), a thin gray font (illegible on a phone), or no card at all.

19 (Add the audio proof). Placement reasoning matters: a sync sound works best at a moment the picture already draws the eye (the hero shot's action, a strong cut) so sound and image reinforce. Strong answers place it early (to prove audio before the viewer decides) or at a peak. The lesson: even a music reel must prove you handle sound — half of every job.

21 (Fix the site). Every gate and its fix: splash animation → remove it, reel first; menu maze → one page; "Work" page burying the reel → reel above the fold; downloadable file → embed to play inline; login wall on client pieces → remove for the public reel (password only genuinely private client work). Meta-lesson: every gate is a place a hirer leaves.

23 (The mobile audit). A three-item fix list from a real phone test: e.g., "reel took 6s to load → compress/host better"; "contact email isn't tappable → make it a mailto link"; "niche text too small to read → increase size/contrast." The point: most hirers are on a phone, so a desktop-only-tested portfolio is broken for the audience that matters.

24 (Footage tells). The commonality among their six proudest shots is the niche signal. Strong answers name a specific through-line ("all six are people talking honestly, well-lit" → founder/nonprofit interviews; "all six are texture and food" → restaurants). The lesson from §39.5: you often notice your niche in your best work rather than choosing it abstractly.

25 (The intersection). Model: they fill all three circles honestly and either find an overlap or identify which circle is thin ("I'm good at and love food work, but I haven't found who pays — I need to grow the market circle by making a spec piece and reaching restaurants"). The value is diagnosing which circle to grow, not forcing a false overlap.

26 (The positioning sentence). Strong: "I make short brand films for independent restaurants that make people hungry and owners proud" — specific enough that a listener can name a business. Weak: "I make videos for businesses that help them grow" — no one can name a referral. The test: read it aloud and ask "who should I meet?" A nameable answer means it works.

27 (Brand consistency check). Grade on whether they found specific inconsistencies across reel, page, and posts — a color clash, a jokey caption over serious work, a values mismatch — and proposed a unification. The lesson: a brand is consistency made visible; inconsistency reads as "no clear point of view," which is what a niche is meant to fix.

28 (The generalist trap). Expected result: the two people can more easily say what they'd hire the specialist for, and often couldn't name a job for the generalist at all. The report should capture the paradox — the "narrower" bio produced more concrete hire-intent. This is §39.5's core argument, felt in miniature.

29 (The tailored ask). Model reply (under 80 words): "Thanks for reaching out! Here's my reel: [link] — the two restaurant pieces around the 0:30 mark are closest to what you're describing. If you tell me your goal and rough timeline, I'll send a quick plan and a price. Happy to jump on a 15-minute call this week if easier." One link, a specific reason to watch a specific part, a clear next step, low friction.

30 (First-portfolio plan). A concrete 30-day plan using the chapter's moves: week 1, reach out to two established shooters to second-shoot; week 2, choose one real local business and shoot a spec piece; week 3, cut and deliver it, ask for a testimonial; week 4, offer one low-rate job for a case study, then set the raised rate. Naming a real business is the tell that they'll actually do it.

31 (Reel specs drill). (a) Portfolio page: 60–90s, 16:9, embedded on a clean player, captioned. (b) Vertical social: 30s, 9:16, safe zones, burned-in captions, faster hook. (c) Corporate producer email: 60–90s 16:9, one link to the reel (not an attachment), a specific reason to watch a specific segment. The skill: match the deliverable to the destination.

32 (Plug the leak). Grade the honesty of the diagnosis. Presence leak → post consistently in the niche, use referrals. Reel leak (found but not converting) → re-cut: stronger hero, harder floor, tighter. Conversation leak (interested but not hiring) → faster replies, clearer scope/price, an easier next step. The one change should target their actual bottleneck, not the stage that's easiest to work on.

33 (Shoot for the reel). The report should notice that shooting with the reel in mind changes behavior — grabbing more details, textures, and small gorgeous moments, holding shots a beat longer, getting a clean version. This is Ch.1's "shoot for the edit" pointed at the reel: the reel is built from inserts, so a reel-minded shooter over-grabs details.

34 (Rule of Six on your reel). For three cuts, name which of Murch's six priorities each serves; a strong answer catches a cut that serves only rhythm and re-cuts it to also serve emotion (e.g., holding a face two frames longer so the feeling lands). The lesson: even in a fast music reel, emotion-first still applies — the best reels make you feel the maker's eye, not just see fast cuts.

35 (Grade for coherence). Guidance: the before/after should visibly unify three separately-graded projects — matched blacks, a shared warmth. The win is realizing that coherence is imposed in post, not required at shoot time; a light unifying grade is what makes disparate work read as one hand, which is what a niche and brand require.

36 (The vertical reel). Beyond the frame change, they had to: raise the floor (30s keeps fewer shots), recompose or drop wides (tall frame punishes them), shorten the hook (social hook is faster — Ch.23), and caption for muted autoplay. The lesson: reframing is a real edit, not a crop — the vertical cut is a harder curation problem, not an easier one.

37 (Deliver and archive). The reel is a deliverable: export a high-quality master (ProRes/DNxHR) and a platform H.264, name it so it's findable in a year (Ch.37), and file it in the project archive. The point: treat your own front door with the same delivery discipline you'd give a paying client's video — you'll re-cut it, and you want the clean source.

38 (Teach it back). A model ~200-word explanation uses the funnel (curate by throwing away more than you keep), the ten-second read (hirers appraise fast), and one example (a beautiful reel with one soft shot reads as "couldn't tell it was soft"). Strong answers land the paradox: a shorter reel of only your best beats a longer one with a weak link, because impression floors at the worst inclusion. Teaching it back is the strongest test of ownership.

39 (The would-I-hire-me audit). Guidance: the value is collecting the pauses — the specific hesitations of three people in or near the niche ("the middle dragged," "I wasn't sure what you specialize in," "I couldn't find your contact"). Those pauses are the next revision list. A student who gets only "looks great!" should push for the harsh version; polite feedback doesn't improve a reel.

Chapter 40 — Answers to Selected Exercises

Model solutions and critiques for the starred (⭐⭐⭐) and odd-numbered exercises, plus items flagged "(Answer/Model answer provided.)". This is the capstone, so many answers are plans with no single right form — judge yours against the reasoning and honesty, not the wording.

A. The whole book, one last time (Synthesis)

1. The four elements, graded. No fixed answer — the deliverable is an honest 1–5 on story/light/sound/edit and a named weakest link. What to look for in your own answer: did the score move from your Chapter 1 self-audit, and can you name which chapter closed the gap? The most useful outcome is naming the still-weakest element (for most readers: sound or story-structure) and turning it into your first "level up" target (§40.3). A grade that's all 5s is not honesty; it's the taste gap hiding — look harder.

3. Trace a decision through the pipeline (⭐⭐⭐). A strong answer keeps one story goal rigid across seven stages. Model (goal = "make the founder feel warm and trustworthy"): pre — schedule the shoot for morning window light and write questions that produce full-sentence, personal answers (Ch.16, 19); camera/lens — a ~50mm-equiv at a wide-ish aperture for gentle background separation, eye-level (Ch.4, 7); light — window as soft key camera-side, gentle fill, no hard shadows (Ch.11, 13); audio — lav close and clean so the voice sounds present and intimate (Ch.14); edit — hold on the eyes after the key line; L-cut the answer over warm B-roll of the work (Ch.28, 30); color — correct to neutral, then a warm, slightly lifted grade (Ch.31, 32); delivery — captioned 1080p master + deliverables so it's clear and reaches everyone (Ch.36). The grade: every choice visibly serves the same feeling. If any stage's choice is generic ("shot it in 4K because that's good"), it failed the trace — the point is that story dictates each one.

5. The Café Scene, then and now (⭐⭐⭐). Success = seven field-by-field lines naming a concrete gained skill + its chapter. Model spine: FRAME → composition/thirds/shot sizes (Ch.6–7); MOVE → the motivated push-in (Ch.8); LIGHT → window-as-key, exposure and fill control (Ch.11, 13); SOUND → clean dialogue + room tone + soundscape (Ch.15, 33); CUT → coverage → assembly → J/L-cuts → story-cut (Ch.26, 28, 29); EFFECT → correction + warm grade (Ch.31, 32); LESSON → "the scene never changed; my command did." The real deliverable is the felt recognition of distance — the same ordinary scene, now fully authored. If a reader can't name the skill for a field, that field is their next study target.

B. The career map (§40.1)

7. Name the rungs. PA (supports the shoot; learns by proximity), operator (runs one craft's gear reliably), DP/editor (authors the look or the cut — the decisions), producer/director (responsible for the whole video: plan, money, people, result). "Reliability" belongs to the operator stage — it's the currency of being re-hired: technical skill becoming dependable execution, take after take, without drama.

9. The non-linear move. Two examples: (a) a producer picks up a camera and operates on a small solo job — they didn't "demote"; they occupy a different box for that job. (b) an editor grows into a DP — moving from authoring the cut to authoring the look — because the underlying question ("what serves the story?") is identical. Why it's allowed: the judgment is one transferable skill — deciding what the video is about and bending every technical choice toward it — so it carries sideways and backward across roles, not just up. The map is a landscape because the skill is portable.

10. The PA's real curriculum (⭐⭐⭐). A strong answer lists decisions to stand next to, not tasks. Model five: (1) why the DP put the key where they did and what they judged by (Ch.11) — watch the light get shaped; (2) why the director asked a question a certain way to get a real answer (Ch.10, 19); (3) why the producer cut a scene from the day's schedule — the money/time trade (Ch.16, 18); (4) how the team solves a problem on the day calmly (Ch.18); (5) what coverage the DP grabs "for the edit" even when it feels boring (Ch.7, 20). The grade: each item is a judgment you're paid to observe, tied to a chapter — turning gear-carrying into the best film school there is.

11. The solo creator's four hats. Model for a 60-second piece: PA — charge batteries, clear the room, control noise before rolling (logistics/fix it in pre); operator — expose with the waveform, frame on a third, get clean audio (execution, Ch.5–7, 14); DP/editor — decide the look (window key, warm grade) and the cut (hold on the reaction, L-cut the B-roll) (Ch.11, 28); producer — decide it's worth making and judge honestly whether it works for a viewer (story is the boss, Ch.1). The point: one small job exercises all four, so the solo creator is already a whole crew.

C. Ways of working (§40.2)

13. Match the temperament. (a) variety + control + tolerates unstable income → freelance; (b) steady paycheck + a mission to belong to → in-house; (c) early, wants to get very good very fast around better makers → agency; (d) wants to build an audience around their own voice → independent creator. Each maps to the column's central trade (freedom vs stability vs fast-learning vs total-control-but-it's-a-business).

15. The business you'd be running (⭐⭐⭐). A correct freelance/creator list includes: finding work, quoting/estimating, contracts/SOWs, invoicing, chasing late payment, taxes/bookkeeping, marketing/outreach, client communication, and admin (Ch.38). The honest realization most people reach: a large fraction of the week — often a third to half early on — is not making video. A strong answer doesn't recoil from that; it either (a) accepts it as the price of freedom, or (b) plans to reduce it (templates, a bookkeeper, batching admin) so more of the week is craft. Judge on honesty about the split and a concrete response to it — not on the exact fraction.

D. Rates, reels, and leveling up (§40.3)

17. The under-pricing autopsy. Too-low pricing is a mistake because: (1) it signals "not serious/hobbyist," so serious clients discount you; (2) it attracts price-shoppers, the exact clients who fight scope, demand endless revisions, and pay late; (3) it's hard to raise later with the same client, who anchored on the low number; and it (4) undervalues the outcome — a testimonial that wins customers is worth far more than "a few hours." A fair price deliberately filters out the price-shopping client who would have hurt you anyway. Humility priced into the invoice is not humility; it's a magnet for the wrong buyer.

19. The reel-selects habit, installed for real (⭐⭐⭐). The deliverable is an actual bin/folder with the best ~10–15s of each of the three projects in it, plus a one-sentence rule (model: "the day I deliver and archive any job, I pull its two best shots into reel-selects before I touch anything else"). Why at delivery beats reel-time: at delivery the footage is online, fresh, and organized and you remember which shot was great; a year later it's on a cold/dead drive, un-remembered, and re-finding it is the "two-day archaeological dig" of Ch.37. Harvest warm; the reel then re-cuts in an afternoon.

21. The rate ladder (⭐⭐⭐). A strong answer shows three rising numbers, each justified by specific new proof. Model shape: today — a starter project rate justified by the reel + one niche spec piece; after 3 paid jobs — meaningfully higher, justified by delivered work, client testimonials, and demonstrated reliability; after a year — higher again, justified by a current reel of paid work, referrals, a defined niche, and a new capability (e.g. motion graphics) that lets you take whole jobs solo. The grade: each jump is tied to concrete proof, not to hope — which is exactly what makes "raise your rate" feel like a decision instead of a gamble.

E. AI in video (§40.4)

23. The taste argument (⭐⭐). Model (~120 words): "When cameras learned to auto-expose and auto-focus, the skill of 'getting a technically clean image' stopped being scarce — anyone could. But image making didn't stop mattering; the scarce skill moved from 'can you get a clean image?' to 'do you know what image to get?' AI is the same shift, bigger and faster: it makes generating, cutting, captioning, and reframing cheap, so the piles of competent-but-pointless video grow — and the rare person who knows which option is right, who has a point of view, becomes more valuable, not less. AI commoditizes execution. It cannot supply taste, because taste is knowing which of a thousand generated options actually serves this specific piece. So the more the tools can do, the more the one thing they can't do — judgment — is worth."

25. The AI ethics memo (⭐⭐⭐). A strong ~150-word memo names all five checks and picks one as non-negotiable. Model checks: (1) consent for any real person's likeness — no non-consensual synthetic media (Ch.35 §35.6); (2) rights — is the AI output licensable/ownable enough to sell to this client, and were the tools' terms read? (Ch.38 §38.5); (3) disclosure — the client knows what's AI-generated; (4) audience honesty — nothing fabricated is passed off as real footage of real events; (5) hand-correction — captions and outputs reviewed, not shipped raw. The check you must never skip: consent for a real person's likeness — because a convincing non-consensual fake is a genuine harm to a real human and a reputation-and-legally-ending act, whereas the others are (mostly) recoverable mistakes. Accept any well-reasoned pick, but consent/harm-to-a-real-person is the strongest answer.

26. Use one tool, keep the judgment (⭐⭐⭐). No fixed answer — the deliverable is a real AI-assisted task done + a list of judgment overrides. Model (auto-caption): the tool typed the words fast; you had to fix the misspelled name, add the missing punctuation, correct a technical term it garbled, fix a line spoken over noise, and re-time one caption that lingered. Model (auto-reframe to vertical): the tool tracked the wrong subject in the two-shot; you re-keyframed to hold the speaker and kept the caption clear of the safe zone (Ch.23). The learning to state explicitly: the machine executed; you decided and took responsibility — which is the whole chapter.

F. What changes and what won't (§40.5)

27. Two layers. Moving layer: a new codec; a higher-resolution sensor; a new platform aspect ratio; a faster AI editing tool. Holding layer: knowing what your video is about; leading the viewer's eye; getting a person to trust you on camera; the fact that a cut makes meaning. The test: does it decide what the video should be (holding) or how a step is executed (moving)?

29. The century-old techniques (⭐⭐⭐). A strong answer picks three durable principles and roots each outside film. Model: (1) leading the eye with composition and light → from painting (centuries of directing a viewer's gaze); survives because human vision, not technology, is what it exploits. (2) story making a stranger care → from oral storytelling, the oldest human technology; survives because it's about human attention and empathy, not tools. (3) the cut makes meaning → demonstrated a century ago (Kuleshov, Ch.1) and rooted in how the mind assembles juxtaposed images; survives because it's a fact about cognition. The through-line: each principle exploits something about humans (vision, attention, cognition) that no tool changes — which is exactly why they outlast every camera and model.

G. Your next twelve months (Plan)

30. Write the twelve-month plan (⭐⭐⭐). No fixed answer — this is the Production Checkpoint. A strong plan has: a niche in one sentence that's specific and referable; a findable reel home; a real rate priced to value; twelve monthly ship-goals, one deliverable each, sized to real available hours; three named prospects + a first move this week; and a chosen path with an honest reason. Weak plans are vague ("get better," "post more"), gearless-goal-free, or have empty months. The strongest test: could the reader start month one tomorrow from this document? If not, it's a wish, not a plan. Require the first move done, not just written.

31. Make the first move this week (⭐⭐⭐). The deliverable is action + a report. Model critique of the outreach message: a strong one is two sentences, about the prospect's need, reel linked — e.g. "Your customer reviews are excellent — a 60-second testimonial video would put that trust right on your homepage and your socials. Here's a 75-second reel of similar work I've made: [link] — happy to send a quick idea for yours." A weak one is long, about the sender ("I'm a passionate videographer who loves storytelling…"), with no specific reason and a buried link. Judge on: was it about them, was it short, did the reel open with the best 5 seconds and load fast? No reply from all three is the normal starting rate — iterate and send three more; the lesson is stamina, not charm.

33. The audience letter, career edition. Model three sentences (client, +12 months): "You made us a testimonial video that we put on our homepage and it's the first thing new customers mention. It made our little shop look as trustworthy as we actually are, and bookings went up after we posted it. When my friend needed video for her business, you were the only person I thought to send her to." The value of the exercise: designing the outcome and the referral you want makes the work aim at it — the most professional possible way to begin.

H. Interleaved and send-off

35. Close the loop with Chapter 1 (⭐⭐⭐). No fixed answer — a strong ~200-word letter to your Chapter-1 self makes three moves: (1) names what you now know that they didn't (that the craft was never the camera — it's the decision, repeated through every frame/light/cut); (2) names a real surprise from the journey (commonly: how much sound mattered, or how much the edit — not the shoot — made the video, or how the quiet workflow chapters saved you); (3) gives the one thing to hold on to when it got hard (keep shipping badly-then-better; the disappointment is the taste gap closing, not a verdict). Model gist: "You think this is about learning a camera. It isn't. In forty chapters you'll stop asking 'what should I buy?' and start asking 'what am I trying to say?' — and that question, which is free, turns out to be the whole job. The thing that will surprise you most is that your worst enemy was never gear; it was waiting to feel ready. You never will. Ship anyway. The first videos will embarrass you — that embarrassment is proof your taste is ahead of your hands, exactly as it's supposed to be, and the only cure is the next one. Hold on to this: the café is still there, the window still lights it for free, and the only variable that ever mattered was you." The grade: does it prove, in the writing, that the reader now owns the book's central idea? That ownership is the truest measure of the distance traveled — and the real final deliverable of the whole book.