Case Study 2: Shoot-Along — A Clean Key, From Green Cloth to Finished Composite
This is a production walkthrough, not an analysis. We light and shoot a green-screen key from scratch in the talking-head setup, then pull, clean, composite, and grade it — every decision shown, every mistake fixed on the day. It uses only gear a phone-first reader can assemble, and it directly builds your Project 3. Where Case Study 1 studied an invisible composite, this one makes one.
The brief and the constraints
Project 3 is a 5-minute branded piece. The brief calls for a spokesperson — the founder of a small coffee-roasting business — to speak to camera from inside a clean, modern, softly-lit studio with the product on a plinth beside them and the brand's warm identity in the background. There is exactly one problem: that studio does not exist, and renting one is not in the budget. What we have is a spare room, a few dollars of green fabric, two small LED panels, a phone, and a clean background plate — a licensed still of a bright, minimal branded space that matches the brand's look.
So the composite is motivated before we touch a camera, which is the only good reason to do one: the story needs the spokesperson in a branded environment we can't build. We will shoot them on green, key them out, and place them into the plate so convincingly that no one watching thinks "green screen" — they think "nice studio."
The constraints, stated plainly, are the ones a phone-first reader actually faces:
- No studio, no cyclorama — a green cloth on a wall, gotten as flat as we can.
- Two soft lights for the screen, one lamp and a bounce for the subject, one small light for a rim.
- A phone shooting its highest-quality, flattest mode.
- A known plate, chosen first, so we light the subject to match a real target rather than guessing.
That last constraint is the secret of the whole shoot. Compositing two things that were never meant to meet is hard; shooting a subject to a plate you're already looking at is easy. We pick the background before we light the person.
⚙️ Settings Box: the green-screen talking-head shoot (our actual settings — adapt to yours).
text PLATE (chosen first) bright, soft, slightly warm branded room; key light reads from the LEFT SCREEN matte green fabric, steamed and clamped taut on a spare wall SCREEN LIGHT two soft LED panels, ~45° each side, EQUAL power, flat even wash, no hotspots SUBJECT LIGHT soft key camera-LEFT (matches the plate) + white bounce fill camera-right RIM one small hard light behind/above the subject → clean edge that beats green spill SEPARATION subject ~7 ft in front of the screen (kills spill, drops shadow off the screen) CAMERA phone, locked on a tripod, ~50mm-equiv feel, eye level (matches the plate's height) RESOLUTION/FPS 4K, 24 fps (matches the rest of Project 3) SHUTTER 1/50 s (180° rule, Ch.2) — motion blur to match the plate/edit EXPOSURE screen = a clean, even mid-green (NOT blown); subject exposed for the face (Ch.5) WHITE BALANCE set to the subject light and LOCKED — no auto drift CODEC the flattest, least-compressed mode the phone offers; expose carefully to protect it AUDIO lav on the subject, close and hidden (Ch.14); record 30 s room tone ALSO GRAB a CLEAN PLATE of the empty green frame + a few seconds of the subject NOT talking
The setup: two lighting jobs, drawn
Before we roll, the plan on paper. The whole shoot is the two separate lighting jobs of §35.2 — light the screen flat, light the subject for the plate — plus the distance that keeps them apart. Here is the room, top-down:
FIGURE CS2.1 — Our green-screen talking-head, top-down (the two-jobs setup)
[========== GREEN CLOTH — steamed, clamped taut to the wall ==========]
↑ ↑
│ even soft wash, both sides, equal power │
▢ ─┘ (screen panels aimed ONLY at the cloth) └─ ▢
(screen L, soft) (screen R, soft)
←──────── ~7 ft separation ────────→
(kills spill; drops the shadow off the cloth)
▣ rim (small hard light, high/behind)
╲ → clean bright edge that overpowers green spill
▼
( S ) ──────► eyeline just beside the lens (to camera, Ch.19)
↗ ↖
▢ KEY ○ white bounce (fill)
soft, camera-LEFT camera-right, ~1.5 stops down
(MATCHES the plate) (matches the plate's gentle contrast)
│
((• lav on subject (hidden), + 30 s room tone
│
[CAM] phone, locked off, eye level, 50mm-equiv, 1/50 s
Two jobs: (1) SCREEN — flat, even, one green value. (2) SUBJECT — lit to belong in the plate,
with a rim to beat spill. The 7 ft between them is the free tool doing most of the work.
And the frame we're composing for — note we frame the subject exactly where they'll sit in the final plate, leaving room for the product and the background identity:
FIGURE CS2.2 — Framing overlay: compose for the FINAL composite, not the green room
16:9 frame O = thirds intersections; S = subject
┌───────────────O───────────────────────O───────────┐
│ │ • subject on the RIGHT third
│ (plate: branded space + product here) │ (leaves left space for the
│ S ← head near upper third │ branded room + product in
│ /|\ correct headroom │ the plate)
│ O / | \ O │ • eyeline to camera (Ch.19)
│ (product plinth will sit here) │ • generous around the subject:
│ │ don't crop where the plate
└────────────────────────────────────────────────────┘ needs to breathe
Compose the GREEN shot for the composite you're building — frame the person where they'll live
in the plate, with room for everything the background will add.
Composing for the composite is the step beginners skip: they frame a nice shot of a person on green, then discover in post that the subject is centered right where the plate's product is supposed to go. We frame the green shot for the final image.
The shoot, in phases
Phase 1 — Light the screen flat
We hang the cloth and fight the wrinkles first, because every wrinkle is a shadow and every shadow is a second green value. Steam it, pull it taut, clamp the edges. Then the two soft panels, one each side at 45 degrees, equal power, aimed to wash the cloth evenly — and we check, not guess: we look at the green on the phone and, ideally, on a waveform (Chapter 5), hunting for hotspots and dark corners. First pass: the center is a stop brighter than the corners. We pull the panels back and feather them (aim them slightly past the screen so the softer edge of each beam covers the center) until the green reads as close to one even value as we can get.
FIGURE CS2.3 — Described Shot: the lit screen (Phase 1 result) [constructed teaching example]
THE FRAME The empty green cloth, filling the background, evenly lit corner to corner — no hot center,
no dark edges, no wrinkle shadows. Boring, flat, uniform. Exactly what we want.
THE MOVE Locked off; the camera won't move all shoot.
THE LIGHT Two soft panels, 45° each side, equal power, feathered to even the wash. Screen light only.
THE SOUND (none yet — lighting pass)
THE CUT —
THE EFFECT A single, clean green value: one easy color for the keyer to find later.
THE LESSON You cannot key what you didn't light evenly. A flat, uniform screen is the whole game;
everything in post gets easier or harder right here.
WHAT WE ADJUSTED Pulled the panels back and feathered them to kill a bright center; re-clamped a wrinkle.
Phase 2 — Light the subject for the plate
Now the other job. We put the subject in their mark, seven feet forward, and we forget the screen entirely — we light them for the plate. The plate's key reads soft and from the left, gently warm, low contrast (a bright, friendly branded room). So: a soft key camera-left, a white bounce camera-right about a stop and a half down to match the plate's gentle ratio (Chapter 11), warm-ish white balance locked to it. We are not lighting a person on green; we are lighting a person who will stand in that room.
FIGURE CS2.4 — Described Shot: the subject, lit for the plate (Phase 2) [constructed teaching example]
THE FRAME Medium shot, subject on the right third, product plinth space to their left, correct headroom.
THE MOVE Locked off.
THE LIGHT Soft key camera-left (matches the plate), gentle bounce fill camera-right ~1.5 stops down.
Warm-ish, low contrast — a friendly, trustworthy studio look.
THE SOUND Lav on, close and hidden; levels checked on headphones (Ch.14); room tone to follow.
THE CUT Will cut against screen-replacement inserts of the product's app — must match them.
THE EFFECT The person already looks like they belong in a bright branded room — before any compositing.
THE LESSON Match the subject's light to the PLATE, not the screen. The composite is half-won here.
WHAT WE ADJUSTED Warmed the key a touch and lowered the fill to hit the plate's soft, low-contrast look.
Phase 3 — Separation and rim: beat the spill
We check the subject's edges against the green. Even at seven feet there's a faint green kick on the shoulders and hair — spill. Two moves: we add the small rim light high and behind, throwing a clean, slightly warm edge that overpowers the green fringe, and we make sure the screen panels aren't wrapping onto the subject (they aren't, thanks to distance). We look again: the green rim is mostly gone, replaced by a clean edge that also separates the subject from the background — the exact job a rim light does in three-point lighting (Chapter 11), now doing double duty as spill insurance.
FIGURE CS2.5 — Described Shot: separation and rim (Phase 3) [constructed teaching example]
THE FRAME Same medium shot; now a clean, subtle bright edge traces the subject's hair and shoulder.
THE MOVE Locked off.
THE LIGHT Key + bounce fill (from Phase 2) + a small hard RIM high/behind. Screen wash unchanged.
THE SOUND Lav rolling; we'll grab room tone at the end.
THE CUT —
THE EFFECT The green spill is overpowered by a clean rim; the subject reads as separated and solid.
THE LESSON Distance + a rim light beat spill on set. Every bit of green you kill here is green the
keyer won't have to eat out of the subject's edges in post.
WHAT WE ADJUSTED Nudged the rim brighter until the last green edge on the hair read clean, not green.
Phase 4 — Roll takes, then grab the safety shots
Only now do we shoot performance: several takes of the spokesperson delivering to camera, eyeline just beside the lens (Chapter 19), directed for a relaxed, warm read (Chapter 10). Then the shots beginners forget and post begs for:
- A clean plate of the empty green frame — the lit screen with no one in it. If the keyer struggles in a corner, this is a reference for what "pure green" looks like there.
- A few seconds of the subject holding still, not talking — a stable reference to set the key on before applying it to the moving performance.
- 30 seconds of room tone (Chapter 15), because sound sells a composited space as much as picture, and the mix (Chapter 33) will need it.
The problem we almost shipped
Halfway through the takes, we notice it on the monitor: the founder's shirt has a subtle sage-green weave, and the keyer preview is starting to nibble a hole in their shoulder. This is the single most common green-screen trap — the subject is wearing the key color — and it is far cheaper to catch on set than to roto out later. We have two honest options: change the wardrobe, or, since the shirt is close to the screen's green, we could switch the whole shoot to a blue screen (§35.2 exists for exactly this). Wardrobe is faster: the founder has a warm brown jacket in the car that suits the brand better anyway. We swap it, re-check the preview — the hole is gone — and re-shoot the takes we'd already banked.
This is worth dwelling on because it is the chapter's premise in miniature. Ten seconds of looking at a rough key on set saved hours of hand-roto in post, and cost only a wardrobe change. Had we "fixed it in post," we'd have been masking a moving shoulder frame by frame for a problem a jacket solved instantly. Always pull a rough key on the day if you can — the monitor tells you the truth while you can still act on it.
Wrap. Everything the composite needs is in the can, and — the chapter's whole point — the hard part is already done. A clean, even screen, a subject lit to the plate, spill beaten with distance and a rim, a good codec, and no wardrobe fighting the key. The key is going to be easy, because we earned it on set.
The post pass: from green to gone
Ingest and prep
Offload and back up (Chapter 27), drop the takes on the timeline, and — because keyed, composited shots are heavy — generate proxies so the timeline stays responsive. Pick the best performance take. Now we build the composite, in Resolve's Fusion page (the free tool this book anchors to; every editor has an equivalent keyer and compositor — Appendix E translates the node names).
Pull and clean the key
We run the exact five-step order from §35.3, and we watch the matte in black-and-white the whole time.
FIGURE CS2.6 — Our node chain (signal flows left → right; Resolve's Fusion page)
[ MediaIn: ] [ KEYER ] [ DESPILL ] [ MATTE ] [ MERGE ] [ result ]
[ subject ] → [ sample ] → [ kill the ] → [ clip black, ] → [ subject ] → over the
[ on green ] [ green → ] [ green edge] [ clip white, ] [ OVER the ] branded
[ (best take)] [ make matte] [ shrink+soft ] [ plate ] plate
▲
▲ garbage matte (a mask) crops a light-stand │ [ MediaIn: branded plate ]
edge out of frame BEFORE the keyer runs └──( the room we're building )
Reading the matte in B/W at each step:
Step 1 (pull): background near-black, subject solid white, hair a clean thin grey — good, thanks to
an even screen. A faint grey haze in one corner → the garbage matte crops it out.
Step 3 (despill): the last green on the hair edge neutralizes; the edge stops glowing.
Step 4 (clean): clip the near-black to pure black; clip the near-white to pure white; shrink the matte
ONE pixel and add a hair of edge blur — enough to hide fringe, not enough to eat hair.
Because the screen was even and the codec was clean, this takes minutes, not hours. The one real fix is the corner haze, which the garbage matte handles, and a touch of despill on the hair. We resist the urge to over-clean — the moment the fine strands start turning into a hard helmet, we back the shrink and blur off. A hair of softness beats a cut-out edge every time.
Composite and grade together
We lay the keyed subject over the plate — and now judge it in color, against the real background, where a green fringe would show against the bright room. It's clean. Here's the stack we've built:
FIGURE CS2.7 — Our composite, as layers (what the viewer sees vs. what it's made of)
WHAT THE VIEWER SEES = FOREGROUND (keyed) through ALPHA OVER BACKGROUND PLATE
┌──────────────────────┐ ┌───────────────┐ ┌───────────┐ ┌───────────────┐
│ founder in a bright │ │ founder, │ │ WHITE=keep│ │ branded room │
│ branded studio, the │ = │ green removed,│ × │ BLACK=drop│ + │ + product on │
│ product beside them │ │ despilled │ │ GREY=hair │ │ plinth (still│
│ edges clean, one │ │ edge │ │ +edges │ │ we licensed) │
│ light on both │ └───────────────┘ └───────────┘ └───────────────┘
└──────────────────────┘ layer 2 its matte layer 1
THEN — the step that seals it — GRADE THE STACK AS ONE:
• one shared look, contrast, and color temperature over BOTH layers (Ch.32)
• founder's SKIN held on the flesh-tone line (Ch.32 §32.5) — the loudest tell if it drifts
• a touch of matching GRAIN over everything, so the clean subject and the still plate share a texture
• a subtle LIGHT WRAP: a thin band of the room's brightness bleeds onto the subject's edge
Ungraded, it's two images sharing a frame. Graded together, it's one photograph.
The grade is what turns two layers into one photograph. We push both to a single warm, bright look, hold the founder's skin on the vectorscope's flesh-tone line so the warmth doesn't jaundice them (Chapter 32), lay a little matching grain over the whole frame so the clean subject and the still plate share a texture, and add a whisper of light wrap so the room's brightness kisses the subject's edge the way real light would. Now it reads as photographed.
The bonus beat: a screen replacement
The brief also wants a shot of the product's app on a phone. On set we filmed a hand holding the phone with a solid blue card where the screen goes (blue, not green — the founder's shirt was green-ish, and §35.2 says switch colors when the subject wears the key color). In post: corner-pin track the four corners, pin in the app UI, then sell it — a soft glow onto the hand, a faint reflection, a hint of glare (§35.5). It cuts against the composited talking-head, so we match their grade. Two composites, one motivated brief.
The fresh-eyes check
Before we call it done, one discipline that catches what tired eyes miss: we walk away for a few minutes, then watch the segment straight through at speed, once with sound and once without, asking a single question — can I point to the effect? The first pass flags one thing: on a bright frame, a faint pale line traces the founder's shoulder where the light wrap is a hair too strong, reading as a halo rather than as light. We pull the wrap back until it disappears. The second pass, sound off, confirms the composite holds in stillness; sound on, it confirms the room tone and music make the invisible studio feel like a place. Then we spot-check it on a phone screen (most viewers' worst-case display, Chapter 33's rule) to be sure the grade and the edges survive a small, bright panel. Only then is it finished.
The timeline
Here is how the finished beat sits on the timeline, composite and inserts and sound together:
FIGURE CS2.8 — The finished segment on the timeline (video over audio; | = cut)
V2 [ light wrap / grain adjustment layer over the whole segment ...................... ]
V1 [ COMPOSITE: founder over branded plate ][ screen-replace: app insert ][ composite ]
A1 [ founder lav dialogue (cleaned, leveled, Ch.33) .................................. ]
A2 [ room tone bed (from set) ....................................................... ]
A3 [ soft branded music bed (licensed, Ch.38) ...................................... ]
^ the app insert cuts in on the founder's line
"...just open the app and—" (an L-cut: her voice
leads us into the screen-replacement shot, Ch.28)
The composite holds across the cut because every shot shares one grade, one grain, one light — and the
SOUND (clean dialogue, room tone, music) sells the branded "studio" as much as the picture does.
Notice the sound doing its half of the work (Chapter 33): clean dialogue, a room-tone bed so the invisible "studio" has an acoustic space, and a soft licensed music bed. A composited environment with dead, roomless audio feels fake no matter how clean the key — the ear has to believe the room too.
Discussion questions
- We chose the background plate first, then lit the subject to match it. Why is that so much easier than shooting a subject on green and finding a background later? What does it change about the lighting decisions?
- Three on-set moves beat green spill here: distance, a rim light, and keeping the screen's light off the subject. If you had to drop one (small room, few lights), which would you fight hardest to keep, and why?
- The grade — one shared look, protected skin, matching grain, a light wrap — is what "sealed" the composite. Explain why an ungraded composite always reads as fake, even with a flawless key.
- We shot a clean plate, a still reference of the subject, and room tone as "safety" grabs. For your own shoot, which of the three would save you the most trouble in post, and what post problem does each solve?
- The screen replacement used a blue card, not green, because the subject wore green-ish clothing. Give another real situation where you'd switch the key color, and say which color you'd choose.
- Sound "sells the studio as much as the picture." Why does a composited environment need believable room tone and ambience? What happens to the illusion without it?
Your turn: your own motivated composite
Extend this walkthrough into your Project 3.
The brief: produce one motivated composited shot for your branded piece — a green-screen key of a subject into a plate, or a screen replacement — following this case study's method end to end.
Do all of it:
- Choose the plate/target first. Pick the background (or the replacement UI) before you light anything, and note its light direction, color, hardness, and contrast.
- Two lighting jobs. Light the screen flat and even; light the subject to match the plate; put 6–10 ft between them; add a rim to beat spill. (Screen-replacement path: light the scene normally, put a clean trackable color where the screen content goes, keep the shot fairly steady.)
- Grab the safety shots. Clean plate, a still reference, and room tone. Always.
- Key or track in five steps. Pull → read the matte in B/W → despill → clean gently → composite. (Or: track the four corners → pin the replacement → sell it with glow/reflection/glare.)
- Grade the stack as one. One look, protected skin, matching grain, a touch of light wrap. Then add the sound that makes the space believable.
- Judge it by disappearance. Show it to someone and ask, "Where were they actually standing?" or "What's on the phone?" If they answer without spotting the effect, you've built an invisible composite — and finished the last creative layer of Project 3.
Key takeaways
- Motivate it first: we composited only because the story needed a branded studio the budget couldn't build. No motivation, no composite.
- Choose the plate first, then light to it. Shooting a subject to a known target is far easier than marrying two images that were never meant to meet.
- Two lighting jobs, kept apart by distance: a flat, even screen (one green value) and a subject lit for the plate, with a rim to beat spill. This is where the key is really won.
- Grab the safety shots — clean plate, still reference, room tone — the ten-second grabs that save hours and sell the space.
- Key in the five-step order, judging the matte in black-and-white, cleaning gently enough to keep the hair. A clean shoot makes this minutes of work.
- Grade the stack as one — shared look, protected skin, matching grain, a light wrap — because the grade is what turns two layers into one photograph. And let the sound build the room the picture only implies.
- Every step runs on a phone and free software. The composite's invisibility is a discipline, not a budget — the same lesson Case Study 1 proved at the top of the industry, now in your own spare room.