Case Study 2: Working the Hinges — Turning a Straight-Cut Assembly into a Scene
This is a from-scratch edit walkthrough — a constructed teaching example (Tier 3), illustrative in its subject but real in its process. Where Case Study 1 analyzed a finished film, this one rebuilds a scene at the timeline, cut by cut, so you can watch a correct-but-flat assembly become a scene that flows. Everything here is doable in any editor that lets you trim audio and video separately (DaVinci Resolve is the book's reference; Appendix E translates every step). Follow the thinking, not just the steps.
The scene and the footage you have
You're deep into Project 2, your 3-minute documentary short. Its subject — our illustrative one — is a furniture maker who builds chairs by hand in a small workshop. You shot it the way Part IV taught you: a sit-down interview (Chapter 19) where she talks about learning the craft from her father, and a generous pile of B-roll (Chapter 20) of her working — measuring a board, sawing, planing, fitting a joint, running a hand along a finished seat. You covered the sawing in two sizes (a wide of her whole body at the bench and a close-up of the blade in the cut), because Chapter 9 told you to shoot the overlap so you could cut on the action later. And, because Chapter 15 nagged you into it, you recorded thirty seconds of workshop room tone and a few minutes of raw tool sounds — the saw, the plane, the mallet — on their own.
You've already done the Chapter 26 job: pulled selects and built an assembly — the shots in the right order, roughly trimmed, telling the story start to finish. Watch it and everything is there. The story is clear. The shots are good. And it plays like a slideshow: her face, cut, a board, cut, her face, cut, a saw, cut. Correct, and dead. This case study is the pass that fixes it — not by adding footage, but by working the hinges between the shots you already have.
Here is what you're starting from.
FIGURE CS2.1 — The "before": a straight-cut assembly (video above audio; | = a cut)
V1 [ INT: her face ####### ][ B: board ##### ][ INT: face ##### ][ B: saw wide ##### ][ B: saw CU ##### ]
A1 [ her voice: "my father..."][ (silent) ][ voice: "...every joint"][ (silent) ][ (silent) ]
A2 [ (nothing under it) ......................................................................... ]
^every cut is a STRAIGHT cut: picture and sound change together, and the B-roll is SILENT
The problems: (1) B-roll drops in with no audio — the sound "holes out" every time we leave her
face. (2) Picture and sound snap together on every cut — it clicks. (3) No bed ties it together.
Three faults, and all three are hinge faults, not footage faults. The B-roll is silent, so the sound falls into a hole every time you cut away from her face. Every cut is a straight cut, so the edit clicks like a metronome. And there's no continuous sound under any of it, so every seam is exposed. You will fix all three with the grammar of this chapter, in four passes.
⚙️ Settings Box: the split-edit workflow. Starting points to set up before you begin — adjust to your editor (→ Appendix E for non-Resolve software).
Setting Choice Why Audio/video link Unlink (Resolve: turn off Linked Selection) so you can trim an audio edge without dragging the picture Audio tracks at least 3 (A1 sync voice, A2 tool SFX, A3 room-tone bed) split edits and beds need room to overlap Trim tool learn the trim/roll edit and the keyboard nudge (1 frame / 5 frames) split edits are made by nudging edges a few frames Snapping On for laying clips, off while fine-trimming a split snapping helps you place, then gets in the way of feel Playback loop-play across each seam you judge a cut by playing over it, not by staring at it The one habit that matters most: judge every cut by playing across it, not by looking at the timeline. A split edit is felt, not seen.
Pass 1 — L-cuts: give the B-roll a voice
The most urgent fault is the silent B-roll, and the fix is the most valuable cut in nonfiction: the L-cut. Right now, when you cut from her face to the board, her voice stops. It shouldn't. She should keep talking — "my father taught me that every joint has to be tight, no glue hiding a bad cut" — over the shot of the board and the saw. The picture cuts to the work; the voice carries on. The viewer hears the story and sees the proof at the same instant, and the cut to B-roll disappears under the continuing voice.
You do it by extending her interview audio forward, under the B-roll picture. Unlink the audio, grab the tail of her voice clip at the cut, and drag it to run under the board and the saw. The picture cuts where it did; the sound no longer does.
FIGURE CS2.2 — Pass 1: L-cut the interview voice over the B-roll
V1 [ INT: her face ##### ][ B: board ##### ][ B: saw wide ##### ][ B: saw CU ##### ]
A1 [ her voice: "...every joint has to be tight, no glue hiding a bad cut" ........... ]
^ PICTURE cuts to the work ^ her voice finally ends here
Her voice now L-cuts OVER the B-roll: we hear the story and see the proof at once, and the
cut to B-roll vanishes under the continuing voice. This one move fixes the "silent hole" fault.
Play it across the seam. The scene has already changed character: instead of "here's her face / here's a board," it's "here's what she's telling us, illustrated." That is the whole reason B-roll exists — to give an L-cut somewhere to land — and it's why Chapter 20 made you over-shoot it.
✂️ In the Edit: this pass was won on the shoot. The L-cut only works because two things exist: the B-roll (shot per Chapter 20) and a clean, continuous stretch of her voice with no cut in it (recorded per Chapters 14–15). If her audio had a hard edit in the middle — a stumble trimmed out — the voice would jump under the B-roll and the L-cut would fall apart. This is why you record answers in full, clean sentences and grab plenty of cutaways: so the edit has continuous sound to run and pictures to run it under. The shoot fed this pass.
Pass 2 — J-cuts: hear the workshop before you see it
Now the opposite move, to enter the working world with momentum. As she finishes a sentence about her father, you're going to cut to her actually working — and you want the sound of the workshop (the rasp of the saw, the tap of a mallet) to arrive before the picture does. That's a J-cut: the next shot's sound leads its picture in. The saw's sound sneaks in under the tail of her interview shot, and by the time you cut to the blade, the ear is already in the workshop.
You lay the raw saw-sound you recorded (Chapter 15) on track A2 and drag its head back under her interview picture, so we hear the cut coming.
FIGURE CS2.3 — Pass 2: J-cut the tool sound in ahead of the picture
V1 [ INT: her face ############# ][ B: saw CU — blade in the cut ############# ]
A1 [ her voice: "...no glue hiding a bad cut" #### ]
A2 [ saw rasp begins HERE ........................ (continues over the picture) ........... ]
^ we HEAR the saw ^ PICTURE cuts to the blade
The saw's sound J-cuts in under her face, so we cut to the blade on a sound we already hear —
the workshop arrives at the ear first, and the picture change lands as a confirmation, not a jolt.
Play across it. The cut into the work now feels inevitable — the ear walked you in. Between Pass 1 and Pass 2 you've built a little rule you'll use forever: L-cut out of the talking (voice lingers over the work), J-cut into the work (its sound leads the picture). The interview and the workshop now interleave like a conversation instead of alternating like flashcards.
🔗 Connection: the two-mic safety and the tool sounds. You could lay the tool sounds because you recorded them separately and cleanly — the "wild sound" habit from Chapters 15 and 19. If you only had the tool sounds buried under interview audio, you couldn't isolate them for a J-cut. Recording clean, separate elements on the day is what gives the edit its freedom.
Pass 3 — cutting on action: hide the saw cut inside the stroke
Here's the fault you haven't fixed yet: the two saw shots — the wide of her whole body and the close-up of the blade — currently cut on a still frame, and when you play it, the saw appears to "jump," because the blade is at a slightly different point in the stroke on each side of the cut. It's a small jump cut (Chapter 9), and it's distracting.
The fix is cutting on action. You place the cut in the middle of a saw stroke, matching the blade's position and direction across the seam, so the motion of the stroke carries the eye across and hides the cut. The saw is moving; the eye is locked on the moving blade; you switch from the wide to the close-up on a frame where the blade is mid-travel — and the jump vanishes into the stroke.
FIGURE CS2.4 — Pass 3: cut on the saw stroke (the motion hides the seam)
THE ACTION (one stroke, two sizes): push ───────► the blade travels ───────► pull
WIDE [ she leans in, arm driving the saw forward ##########| ]
CU [ |## blade mid-cut, travelling ## ]
^ CUT here, mid-stroke — blade moving,
same direction, matched position
Cut DURING the stroke, not between strokes. The moving blade carries the eye across; the
jump disappears. Trim a few frames either way and loop-play until the stroke looks continuous.
You find the exact frame by nudging the cut a few frames at a time and looping across it. Too early and the blade repeats a bit of its travel; too late and it skips ahead; just right and it looks like one continuous stroke seen from two distances. A couple of repeated frames often reads smoother than a mathematically perfect match, because the eye wants continuous motion more than it wants a perfect position — so trust your eye over the numbers, and let the loop-play be the judge.
✂️ In the Edit: the overlap you shot in Chapter 9. You can only cut on the stroke because you shot the whole stroke in both sizes — the overlap. If the wide had stopped the instant before you cut and the close-up started the instant after, there'd be no shared motion to hide the seam, and you'd be stuck with the jump. On the day, running the saw stroke fully in each size felt redundant. Right now it's the only reason this cut works. Always shoot the action past where you think you'll cut, in every size.
⚠️ Common Mistake: cutting between the strokes. The instinct is to cut on the "clean" still moment between two saw strokes — it looks like a tidy edit point on the timeline. It's the worst possible place: with nothing moving, the eye is free to catch the change, and any mismatch jumps. Cut where the motion is busiest, not where it's calmest. The stillness you're drawn to is exactly where the seam shows.
Pass 4 — the bed: one continuous sound under everything
Three passes in, the scene flows far better — but if you listen closely, the ambience still ducks and swells as you cut between the quiet interview room and the louder workshop. The last pass glues everything with a single continuous room-tone bed: you lay your thirty seconds of workshop ambience (Chapter 15) on track A3, loop it to run unbroken under the entire scene, at a low level, and now the picture and the sync sounds can cut as much as they like without the background ever jumping. The bed is the floor the whole scene stands on.
FIGURE CS2.5 — Pass 4: one unbroken ambience bed ties every cut together
V1 [ INT face ][ B: board ][ INT face ][ B: saw wide | saw CU ][ B: hand on finished seat ]
A1 [ her sync voice, L-cut across the cutaways ....................................... ]
A2 [ tool SFX, J-cut in ahead of the work ............. .... .......................... ]
A3 [ WORKSHOP ROOM TONE — one continuous low bed, cut to NOTHING ........................ ]
^ because the bed never cuts, every picture cut and every sync cut floats on steady sound
The unbroken bed is the single most important glue in the scene. It's the thirty seconds of
room tone Chapter 15 made you record — finally doing its job.
Now play the whole thing. The scene breathes. Her voice threads over the work, the workshop's sounds lead you into each task, the saw cut is invisible inside its stroke, and a steady workshop hum holds it all on one continuous floor. It is the same shots, in the same order, as the dead assembly you started with. You added no footage and no effects. You worked the hinges.
The setback (and the save)
No edit is clean on the first pass, and yours isn't either — which is worth walking through, because recovering is an editing skill in itself. Reviewing the near-final scene, you catch a problem at the very cut you were proudest of. Your J-cut into the saw is beautiful — but you'd trimmed a long "um" out of the middle of her sentence just before it, and now, right before the workshop sound comes in, her face does a tiny hop: a jump cut in the interview, left behind when you removed the stumble.
You have three ways out, and choosing between them is the real lesson:
- Cover it with an L-cut to B-roll. Cut away to a cutaway (her hands, the board) across the trimmed spot, keeping her voice running underneath. The jump happens off-screen, hidden behind B-roll — the classic fix, and the reason you shoot cutaways.
- Hide it with a cut on action. If there's a gesture or a head-turn at the trim point, cut on that motion so the movement covers the jump — the §28.2 rescue.
- Absorb it into a deliberate jump-cut style. If the piece elsewhere already uses a punchy, jump-cut energy, a single visible jump might fit. Here it doesn't — the scene is warm and flowing — so this option is out.
You take option 1: you drop a two-second cutaway of her hands smoothing the board over the trim, L-cutting her voice across it, and the jump vanishes behind the B-roll while the story never pauses. The note you write in your project file is blunt and permanent: every trim inside a sentence leaves a jump — plan a cutaway or a motion to cover it before you cut the words. This is Chapter 30's whole method previewed: in an interview edit, you build the story in the audio first, then lay B-roll over every seam so the picture never shows the joins.
🔗 Connection: this is what Chapter 30 is built on. The move you just made — trim the words for story, then hide the picture-seam under a B-roll L-cut — is the engine of the entire interview-driven edit, which is Chapter 30. You've just done in miniature what that chapter does across a whole piece. The advanced cuts of this chapter are the tools; Chapter 30 is the method that uses them at scale.
The "after": before and after, side by side
Here is where you started and where you ended, on one page.
FIGURE CS2.6 — Before vs. after (the same shots, the same order)
BEFORE (assembly, all straight cuts):
V1 [ face | board | face | saw wide | saw CU ] ← clicks; B-roll silent; saw jumps
A1 [ voice | silent | voice | silent | silent ] ← sound holes out on every cutaway
A2 (nothing) ← no bed; every seam exposed
AFTER (the same footage, hinges worked):
V1 [ face | board | saw wide||saw CU | hand ] ← saw cut hidden IN the stroke
A1 [ voice L-cut across the cutaways ............ ] ← voice threads over the work
A2 [ tool SFX J-cut in ahead of the picture ...... ] ← workshop arrives at the ear first
A3 [ room-tone bed, one continuous floor ......... ] ← every cut floats on steady sound
Same shots. Same order. No new footage. The difference is entirely in the hinges.
That last line is the whole lesson of the chapter, made concrete: the gulf between an assembly and a scene is not footage or effects or gear — it is the handful of decisions you make at the boundaries between shots. You L-cut out of the talking, J-cut into the work, cut the saw on its stroke, and floated it all on one unbroken bed. A viewer will feel the difference immediately and will never be able to name it. You can name it: you worked the hinges.
Notice, too, that the order of the four passes was not arbitrary, and it's worth stealing as a workflow. You fixed the most damaging fault first — the silent B-roll (Pass 1), which was breaking the sound — before you touched the merely-flat rhythm. You added the J-cuts (Pass 2) only once the voice was carrying, so you could hear how the tool sounds sat against it. You cut the action (Pass 3) as a discrete, self-contained fix that didn't depend on the audio work. And you laid the bed last (Pass 4), because a bed is a finishing glue — laid too early, it just masks problems you should be solving in the picture and the sync tracks. The general principle: fix what's broken before you polish what's merely flat, and lay the continuous bed last, over a scene that already works without it. A bed should be the seal on a good edit, never the spackle over a bad one.
There's a rhythm lesson hiding in here as well, and it points straight at the next chapter. Once the scene flowed, you probably noticed you could feel where it wanted to breathe — a beat to hold on her face before cutting to the saw, a moment to let a finished chair sit in silence. That instinct — how long to hold, when to cut for feeling rather than for smoothness — is what Chapter 29 is about. The grammar you applied here made the scene correct; the pacing you're now itching to adjust is what will make it move someone. Note those instincts; they're the seed of the fine cut.
🎒 Gear Note: everything here is free. Not one pass required anything you don't already own. Unlinking audio from video, trimming a split edit, cutting on action, and laying a room-tone bed are pure decisions, made in the same free editor (DaVinci Resolve, or any app that trims audio and video separately) you used for the straight assembly. The "upgrade" was thinking, not spending. If your app can't unlink audio from video, that's the single feature worth switching apps for.
🎬 On Set (in the edit): the four-pass rebuild. Take a straight-cut assembly of your Project 2 (or any interview-plus-B-roll scene) and run the exact four passes: (1) L-cut the voice over every B-roll shot; (2) J-cut any location or tool sound in ahead of its picture; (3) cut any action across a size change on the motion; (4) float one continuous ambience bed under the whole scene. Constraint: add no new footage — only work the hinges. Self-review: play the before and after for someone and ask which feels "more professional" and whether they can say why. They'll feel it and struggle to name it. That gap is the craft.
Discussion questions
- All four passes fixed the scene without adding a single new shot. What does that tell you about where the value in editing actually lives — and about the phrase "we'll fix it in post"?
- Pass 1 (L-cut) and Pass 2 (J-cut) form a little rule: L-cut out of the talking, J-cut into the work. Why does that specific pattern make an interview-plus-B-roll scene feel like a conversation rather than a slideshow?
- The saw cut only worked because the stroke was shot fully in both sizes (the overlap, Chapter 9). Describe a time you couldn't make a cut work in an edit because the footage wasn't shot for it. What was missing?
- The setback was a jump cut left behind by trimming an "um." Three fixes were offered; the editor chose the B-roll L-cut. When would the motion fix (option 2) be the better choice instead?
- The room-tone bed is described as "the floor the whole scene stands on." In your own words, why does one continuous background sound let the picture cut freely without the sound seeming to jump?
- This whole walkthrough is called "working the hinges." Take that metaphor apart: in what sense is the boundary between two shots a hinge rather than a wall, and how does thinking of it that way change how you edit?
Your turn: rebuild one scene, hinge by hinge
Take one flat, straight-cut scene from any of your projects and rebuild it exactly as this case study did.
The brief: turn a correct-but-flat assembly into a scene that flows, using only the grammar of this chapter — no new footage, no effects.
The passes:
- Pass 1 — L-cut the voice over the cutaways. Extend your interview or narration audio under every B-roll shot so the sound never holes out. (If you have no B-roll for a seam, note it — that's a shooting lesson for next time.)
- Pass 2 — J-cut sound in ahead of picture. Let a location's or an action's sound arrive a beat before you cut to it, so the ear leads the eye in.
- Pass 3 — cut on the action. Find any movement that crosses a size or angle change and place the cut mid-motion, matching position and screen direction, to hide the seam.
- Pass 4 — float a bed. Lay one continuous ambience or music bed under the whole scene so every cut stands on steady sound.
- Then handle a setback. Find a jump cut left by a word-trim and cover it with a cutaway L-cut or a motion — and write yourself the shooting note that would have prevented it.
Play the before and the after back to back, ideally for someone else. If they feel the second one is "more professional" but can't say why, you've done exactly what this chapter teaches: you've moved a scene from assembled to edited by working the hinges — which is where footage finally becomes film.
Key takeaways
- The gap between an assembly and a scene is the hinges, not the footage. All four passes here added no new shots and no effects — they worked the boundaries between shots.
- L-cut out of the talking; J-cut into the action. Let the voice linger over the cutaways (so B-roll has a voice) and let location/tool sound lead the picture in (so the ear arrives first). That pattern turns a slideshow into a conversation.
- Cut action on the motion, not between motions. The saw cut vanished inside the stroke because the blade was moving and matched across the seam; cutting on the still frame between strokes is where jumps show.
- One continuous ambience bed is the floor. Because the bed never cuts, the picture and sync sounds can cut freely without the background jumping. It's the room tone Chapter 15 made you record.
- Every advanced cut was won on the shoot. The L-cuts needed clean continuous audio and B-roll; the J-cuts needed separately-recorded tool sounds; the cut on action needed the overlap in both sizes. You shot for the edit, and the edit had somewhere to go.
- Trimming words leaves jumps — cover them. Every trim inside a sentence leaves a picture-seam; hide it under a B-roll L-cut or a motion. This is the engine of the interview edit (Chapter 30) in miniature.
- It's all free. Unlinking audio, trimming split edits, cutting on action, and laying a bed are decisions made in the free editor you already have. The upgrade is in your head.