Case Study 2: The Last Cobbler on the Block — Building a Short Doc From Scratch
Where Case Study 1 took apart a landmark film, this one builds a short documentary from nothing — one interview, a morning of B-roll, and an edit that discovers a story. Follow along, and better, shoot your own version. This is a shoot-along-and-edit in the talking-head setup (§3 of the style bible): the profile you can make on a phone, with a real person and a real trade.
The brief
Make a two-to-three-minute short documentary profile of one person with a hands-on craft — here, a neighborhood shoe repairer, a cobbler, working a trade that is quietly disappearing. It must have a real reality arc: a hook, a spine question, a turn, and an ending that earns its feeling. One subject, one interview, a morning of B-roll, natural light where possible, a clip-on mic, and free editing software. We will build it two ways — with narration and without — and decide which the story needs.
We choose a cobbler for the reasons Chapter 1 gave us: there are no boring subjects, only boring choices. A man mending shoes at a bench is ordinary, which is exactly the test. If we can make three minutes of shoe repair carry a real story, we can profile anyone. And a vanishing trade hands us a built-in arc — a question the reality itself is asking: what happens when the last person who knows how to do something decides whether to stop?
(No one in this walkthrough is a named character — we refer to "the cobbler," "our subject," "the repairer." The person is real in the sense that this is exactly how you'd shoot a real one; the specific figure here is a constructed teaching example.)
Pre-production: finding the arc before the shoot
Even a three-minute doc earns real planning, because — unlike a scripted piece — we are hunting for a story we can't fully see yet. The planning is a hypothesis about the arc (§21.2), which reality will correct.
The hypothesis arc. Before we shoot, we write one sentence of the story we think is there: "A cobbler in his seventies works a trade almost no one learns anymore — and has to decide whether it ends with him." We will very likely be wrong in the details. But shooting toward a hypothesis (§21.2) is how we get coverage that can become a story, rather than a drive full of pretty, useless shots.
The spine question. From the hypothesis we extract the question the film will pose and answer: Will this craft outlive him — and does he want it to? Every beat we shoot should push toward that question. It is our filter on set: if a shot doesn't serve the question, we probably don't need it.
The interview plan (Chapter 19). We plan questions designed to produce full-sentence answers that hunt for the want and the obstacle — the arc ingredients, not facts:
FIGURE CS2.1 — Interview question plan (hunting for the arc, not facts)
WARM-UP "Walk me through the first thing you do when you open up." (gets them talking, relaxed)
WANT "What do you love about a shoe most people would throw away?" (the devotion)
CRAFT "Show me something you can do that a machine can't." (the expertise, sets up B-roll)
OBSTACLE "Who's learning this now?" (the stakes — likely: nobody)
THE TURN "What happens to all of this when you stop?" (the emotional core)
BUTTON "Why are you still here every morning?" (the resonant ending line)
Rule (Ch.19): ask open, then quiet. Let silence pull the real answer. Ask them to answer in
full sentences so each bite stands alone when we lift it for the spine.
The B-roll list (Chapter 20). We plan the pictures that will show what the interview says — hands, tools, leather, the shop, the street — as sequences (wide/medium/detail) so the edit can cover every cut:
FIGURE CS2.2 — B-roll shot list (matched to the story beats)
1. EXTERIOR the shop front, the hand-painted sign, the street establish place / mood
2. WIDE the whole shop interior, him at the bench establish the world
3. SEQUENCE hands cutting leather: wide -> medium -> extreme detail the craft (the heart)
4. DETAIL worn tools, a wall of lasts, an old machine texture, history
5. SEQUENCE a finished shoe: rough -> repaired transformation / payoff
6. DETAIL the empty chair, the closed shop next door the stakes (visual)
7. ARCHIVAL an old photo of the shop decades ago (if he has one) the past (§21.4)
+ ROOM TONE 30s of the shop's ambient sound the sound bed (Ch.15)
Notice we plan more than three minutes needs. Coverage is insurance (Chapter 20); we would rather discard footage than discover a hole in the edit.
Gear and settings
The whole point is that this needs almost nothing — a talking-head profile is the most phone-friendly documentary there is.
⚙️ Settings Box: a one-person short-doc shoot (phone or camera).
Setting Value Why Resolution / frame rate 1080p or 4K, 24 or 30 fps Standard doc look; 4K lets you punch in for a second "angle" in the edit Interview audio Clip-on lav or a shotgun close and off-frame Sound is half the picture; a talking head lives on clean dialogue Interview light A window as motivated key (Ch. 13), off to one side Free, soft, flattering; side light models the face Eyeline Subject looks at you, just off the lens (Ch. 19) The off-axis interview look; not straight down the barrel B-roll Same camera; brace or handheld with intent Match the interview; steady for detail, motivated if moving Backup Record room tone; shoot extra seconds each side of every B-roll Patch material and handles for the edit A phone, a clip-on mic, a window, and something to brace on is a complete short-doc kit. Everything else is refinement.
The kit: phone, a ten-dollar lav mic, a small tripod or a stack of books, and the shop's own window light. That's it.
The setup
We seat the cobbler at his own bench, turned so the shop's front window becomes his key light, and we place ourselves — the interviewer — just to one side of the lens so his eyeline is natural and off-axis (Chapter 19). A top-down map:
FIGURE CS2.3 — Interview setup at the workbench (top-down view)
[ SHOP WINDOW ] ☀ soft daylight (the motivated key, Ch.13)
|
▼
( S ) cobbler at his bench ──────► eyeline to interviewer (just left of lens)
↗ ↖
(window is ○ white card / bounce, camera-right,
the key) lifts the shadow side ~1.5 stops
|
((• lav clipped to his collar (or shotgun just out of frame top)
|
[ CAM ] ~50mm-equiv feel, subject on the left third, window side
Key-to-fill ~1.5 stops -> gentle modeling, a natural working-light look, nothing bought.
The single most important choice: the window is to the side, not behind him. Side light reveals the texture of his face and hands — the whole point of a craft profile — and models him in three dimensions. Window behind him would have made a silhouette (the Chapter 1 mistake). We frame him on the left third with the bright window on his key side, leaving him room to look across the frame toward us. The bench, the tools, and the shop fall away softly behind — the world he lives in, present but not competing.
The shoot, phase by phase
We work through the interview first (while he's fresh), then the B-roll — and, crucially, we listen as we go, because the real arc will reveal itself in what he actually says.
Phase 1 — the interview
We mic him, set the window key, and start with the warm-up question while he keeps working — hands busy, nerves settling. Then we move through the plan, asking open questions and going quiet after each answer (Chapter 19), letting silence pull the real thing out.
FIGURE CS2.4 — The interview, the turn [constructed teaching example]
THE FRAME Medium close-up, subject on the left third, looking camera-right to us just off the lens;
soft window key on his near cheek, the shop bokeh'd behind at a working aperture.
THE MOVE Locked off on the little tripod. Stillness — he's about to say the hard thing.
THE LIGHT Window key camera-left of him (our right), a card bounce filling the far cheek a stop-plus
down; the shop's practical bulbs glow warm behind, separating him from the background.
THE SOUND His lav audio, close and clean. We asked "What happens to all this when you stop?" and then
said nothing. Four seconds of room tone. Then, quietly, the answer.
THE EFFECT Because the picture is still and the room is bare, his pause becomes the loudest thing in
the shot. We lean in before he even speaks.
THE LESSON The turn is not a question you ask — it's a silence you protect. Hold the camera and hold
your tongue, and the subject will hand you the heart of the film.
What reality corrects: our hypothesis was "a man mourning a dying trade." But when we ask "who's learning this now?", he surprises us — a young neighbor has started coming by on Saturdays. And when we ask "what happens when you stop?", he doesn't mourn; he shrugs and says the work was never his to keep, only to pass on. That is the real arc, and it's better than ours: not a lament, but a quiet handing-off. We would never have found it if we'd stopped listening after the answer we expected. This is §21.2 in the room — you shoot toward a hypothesis and let the truth improve it.
Adjustment: his best answers ran long and buried the gold mid-sentence. We ask two of them again, gently: "Say that again, but start with 'The work was never mine to keep.'" Coaching a subject to answer in a clean, self-contained sentence isn't dishonest — the meaning is entirely his; we're only helping him give us a bite that will stand alone as a spine line (§21.5: shaping, not fabricating).
Phase 2 — the B-roll sequences
Now the pictures that will carry the story over his voice. We work the shot list, shooting each action as a sequence — wide, then medium, then extreme detail — so the edit can build and cut freely (Chapter 20).
FIGURE CS2.5 — B-roll detail: the hands [constructed teaching example]
THE FRAME Extreme close-up: a curved knife pares a sliver of leather; the grain, the nicks in the
blade, the stained fingertips fill the frame. The shoe is the whole world.
THE MOVE Locked off, braced on the bench. The only motion is the work itself.
THE LIGHT The same window, now raking hard across the leather from the side — texture blooms. Backlit
dust hangs in the air. Nobody lit this; we just chose where to stand relative to the window.
THE SOUND The close scrape of blade on leather, the tap of a hammer — the sounds we'll build the
soundscape from (Ch. 33). Recorded clean, close, over room tone.
THE EFFECT A mundane action becomes mesmerizing purely through framing and raking light. Texture is
the story; the hands say "mastery" without a word.
THE LESSON For a craft profile, the detail sequence IS the film. Get impossibly close to the hands and
let the light rake the texture — that's where a trade becomes cinema.
Thinking: over his line "the work was never mine to keep," we want a picture that shows the idea, not the words (§21.3, the narration rule applied to B-roll too). So we shoot the young neighbor's hands next to his on the same shoe — two generations on one repair. That single planned image will become the film's visual thesis. We didn't know we needed it until the interview revealed the arc; because we were still on location, we could grab it. Shoot for the edit — and re-plan B-roll the moment the real story shows up.
Adjustment: the exterior of the shop read flat and dull in midday light. We come back at the end of the day, when low sun hits the sign, and reshoot — a pre-production lesson learned on the day (Chapter 13, golden hour). Fix it in production, because the edit can't add light that wasn't there.
Phase 3 — archival and room tone
He digs out a single old photograph: the shop, decades ago, a younger man at the same bench. Gold. We photograph it big under the window (§21.4 — scan/shoot archival at high resolution so we can move within it), noting that he owns it and has given permission to use it (§21.5, and Chapter 38 for the formal release). Then, before packing up, we record thirty seconds of the shop doing nothing — the hum, a distant street, a clock. It feels pointless. In the edit it will be the continuous bed that lets every picture cut without the sound jumping (Chapter 15). Always record room tone.
⚠️ Common Mistake: shooting the interview and forgetting to shoot the story. A beginner films a great interview and a handful of generic shop shots, then discovers in the edit that they can't show the film's actual point — the handing-off between generations — because they never shot it. The fix is the one this walkthrough models: let the interview reveal the real arc on location, then shoot the specific B-roll that arc needs before you leave. You can't cut to a picture you never took.
The edit pass
We drag everything into a free editor (DaVinci Resolve is the book's reference from Chapter 26; any editor works) and build the story — which, for documentary, means finding it, not just assembling it (§21.2).
The paper edit (a preview of Chapter 30). Before touching the timeline, we listen back to the whole interview and pull the sentences that could carry the film alone — the soundbite spine. We write them on paper in the order that makes an arc:
FIGURE CS2.6 — The paper edit: the soundbite spine (the story in his own words)
HOOK (cold open, over hands working) "...a shoe most people would just throw away."
QUESTION "Nobody really learns this anymore."
DEVELOP (over the craft sequence) "A machine can't feel where the leather wants to give."
DEVELOP (over the shop, the old photo) "My father stood at this same bench."
THE TURN (the protected silence) "The work was never mine to keep. Only to pass on."
RESOLVE (over the two hands, one shoe) "He comes Saturdays now. He's got the hands for it."
BUTTON (over him alone, closing up) "Somebody has to be the one who still knows how."
The spine is entirely HIS words, arranged to reveal the arc reality actually contained.
Read that column and notice: the story is already there, told entirely in the subject's own voice, before we've added a single frame of B-roll or a word of narration. That is the soundbite spine (§21.6), and building it first is the secret of the interview-driven edit.
Assemble and cover. Now we lay the spine on the timeline and cover every cut with B-roll, so the talking head is barely seen and never jumps:
FIGURE CS2.7 — The 2:45 assembly (spine + B-roll over the cuts)
V1 [exterior][ B-roll: hands ][INT][ craft sequence ][ old photo move ][INT: turn][ two hands ][button]
A1 [ soundbite spine (his voice) ...................................................................... ]
A2 [ narration option? ..... (we'll test this) ..... ]
A3 [ room tone bed + tool sounds ...................................................................... ]
0:00 0:20 0:40 0:55 1:40 2:05 2:20 2:35 2:45
^ picture cuts freely; the room-tone bed (A3) runs unbroken so
the sound never "jumps" at an edit (Ch. 15 / L-cut logic, Ch. 28)
Every picture cut is hidden under B-roll or bridged by the continuous room-tone bed. The interview is the spine; we see his face maybe three times, at the moments that matter most (the turn, the button). Everywhere else, we hear him and see the work — the essence of the B-roll-driven narrative you'll formalize in Chapter 30.
The decision: with narration, or without
Here is the Production Checkpoint made real. We build the opening two ways.
Version A — with narration. We write three lean lines of our own voice-over (§21.3), recorded that night in a closet, to bridge what his interview doesn't cover: who he is, where we are, and the fact — which he never states — that his is the last repair shop for miles.
FIGURE CS2.8 — The narrated open (narration fills only the gaps)
PICTURE NARRATION (ours) / SPINE (his)
───────────────────────────────────── ─────────────────────────────────────────
EXTERIOR: the shop at golden hour VO: "This is the last shoe-repair shop for
eleven miles."
B-ROLL: hands paring leather SPINE (him): "...a shoe most people would
just throw away."
WIDE: him at the bench VO: "He's been at this bench for fifty years."
INTERVIEW: to camera SPINE (him): "Nobody really learns this anymore."
Our narration supplies only facts his words can't (the "eleven miles," the "fifty years").
It never describes the picture, and it goes silent the instant his voice can carry the beat.
Version B — without narration. We cut the same opening carried only by his voice, the sounds of the work, and the pictures — no outside narrator at all. The "eleven miles" fact is gone; instead we open cold on the sound of a hammer and his first line, and let the empty shop next door show the stakes the narration would have stated.
Watching both back-to-back, the difference is exactly what §21.3 promised:
- Version A (narrated) is clearer and faster. In fifteen seconds you know who, where, and why it matters. It feels guided, authoritative, a touch journalistic. The "eleven miles" fact lands a punch no image quite delivers.
- Version B (un-narrated) is more intimate and more trusting. It's slower to orient you, but you feel like you discovered the man rather than being introduced to him. His voice, uninterrupted by ours, feels more like his story and less like ours about him.
Neither is "right." For this film — a quiet, personal profile whose whole theme is a private handing-off — we choose mostly Version B, with one line of narration kept: the "eleven miles" fact, because it's the one piece of context no image or interview line supplies, and it raises the stakes early. We cut the other two narration lines; his own voice covers them better. This is the judgment the whole chapter builds toward: narration is connective tissue, used only where the real voices leave a gap — and hearing the two versions is how you learn to feel where that gap actually is.
Refine. We watch it whole and adjust for feeling. The turn — "the work was never mine to keep" — wants air, so we hold the picture and let the room tone breathe two seconds before the line. We tighten a redundant detail shot. We push in slowly on the old photograph (§21.4), motivated: from the young father toward the bench that's still there. Small moves; the film becomes itself in this pass — the documentary-editing truth that the story is finished in the edit, not the shoot.
Export. One high-quality file (export is Chapter 36), with captions burned or attached (§36.5 — the film isn't finished until everyone can follow it) and a simple lower third naming him and the shop (Chapter 34). Under three minutes. A documentary — found, not captured.
What we'd do differently (and why that's the point)
Watch it honestly and there are already next lessons — and naming them is the craft:
- We under-shot the young neighbor. The film's real thesis — the handing-off — turns on him, and we got exactly one usable shot of his hands. We should have grabbed a full sequence the moment the interview revealed his importance. The arc showed up on location; our B-roll didn't fully catch up to it. Next time: when reality changes the story, re-shoot the B-roll to match, before you leave.
- One narration line we kept still leans toward telling. Even "the last shop for eleven miles" borders on stating a stake we could have shown with a longer look at the shuttered shop next door. It's defensible, but it's the film's closest call to over-narrating.
- The turn could be even barer. We hold two seconds of room tone before his key line; a braver cut might hold four. Silence, used deliberately, is the strongest tool we have — and we were a hair timid with it.
None of these is a failure. They are the next film's lessons, surfaced by finishing this one. That is the loop the whole book runs on: make it, watch it honestly, name one fixable thing, make the next one better.
Discussion questions
- Our hypothesis arc ("a man mourning a dying trade") was wrong, and the real arc ("a quiet handing-off") was better. Why is shooting toward a hypothesis still the right move, even when reality corrects it?
- We coached the subject to restart an answer as a clean, self-contained sentence. Argue that this is honest shaping (§21.5) rather than fabrication. Where would the line be?
- The soundbite spine (FIGURE CS2.6) tells the whole story in the subject's own words before any B-roll or narration. Why build the spine first? What does that discipline protect you from?
- We kept exactly one line of narration and cut two. Using §21.3, explain how you decide which gaps a real voice can cover and which genuinely need narration.
- The film's thesis image — two generations' hands on one shoe — existed only because the interview revealed the arc on location. What does this tell you about the relationship between the interview and the B-roll shoot?
- Version A (narrated) was clearer; Version B (un-narrated) was more intimate. For your Project 2, which does your story need more — and how will you know?
Your turn
Shoot your own two-to-three-minute craft profile following this exact process:
- Write a hypothesis arc and a spine question — one sentence each. Expect reality to improve them.
- Plan the interview to hunt for the arc (want, obstacle, turn), not facts — and plan B-roll sequences that will show what the interview says (Chapters 19–20).
- Shoot the interview first, then listen — let the real story surface, then shoot the specific B-roll that story needs before you leave.
- Record archival big and room tone always (§21.4, Chapter 15).
- Paper-edit the soundbite spine in the subject's own words, then cover every cut with B-roll.
- Build it with and without narration, watch both, and keep narration only where a real voice can't reach.
- Watch it back with headphones and write one Frame Log line: what worked, and the one thing you'd do better.
It will be imperfect. Keep it — it's a real documentary, and it's the baseline you'll measure Project 2 against.
Key takeaways
- The arc is a hypothesis you shoot toward and reality improves. Plan the story you think is there, then listen on location and let the truth make it better.
- The interview reveals the film; the B-roll must catch up on the day. When the real arc surfaces, shoot the specific pictures it needs before you pack up — you can't cut to a shot you never took.
- Build the soundbite spine first. The subject's own words, arranged into an arc, are the story — before a single frame of B-roll or word of narration.
- B-roll over every cut turns a talking head into a film: we see his face only at the moments that matter and see the work everywhere else.
- Narration is connective tissue. Build the piece with and without it, feel the difference (clearer vs. more intimate), and keep narration only where a real voice leaves a genuine gap.
- Shaping is not fabricating. Coaching a clean answer, compressing time, choosing the spine — all honest, because the meaning stays the subject's. The honesty test governs every cut.
- You did all of this with a phone, a window, a clip-on mic, and free software. The craft — finding the true story and honoring it — made it a documentary.