# Turn a script scene into a hand-drawn storyboard animatic with sound

Board a scene like a story artist: source-accurate characters, an approved camera plan, one locked master per setup, a new beat in every drawing, cast voices and a ducked score.

Author: Robert Torres

Canonical: https://voyager.so/blog/script-scene-to-storyboard-animatic

Published: 2026-10-19T22:14:00-07:00
Updated: 2026-10-19T22:14:00-07:00

An animatic is only useful if it plays like the finished scene. If the drawings sit on screen like slides, the camera jumps around inside a shot or the voices sound wrong, it is a slideshow, not a plan a director can cut against.

For this example, we boarded the café scene from chapter 1 of Hemingway's *The Sun Also Rises*: Jake, Robert Cohn and Frances at a table in the Café de Versailles at night, a kick under the table, then the walk up the boulevard to the corner kiosk. The result is 23 shots and 86 drawings in 1 minute 24 seconds, with the book's dialogue, ambience, foley and an underscore.

![A hand-drawn pencil storyboard animatic of a 1920s Paris café at night: a wide of three people at a table, matched singles of Jake and Cohn, a kick under the table, then the two men walking up a lamplit boulevard to a corner kiosk, each drawing labelled with its shot and setup.](https://cdn.moda.app/voyager/blog/script-scene-to-storyboard-animatic/hero-8587121cdb8d24bb.webp)

[Video: A hand-drawn pencil storyboard animatic of a 1920s Paris café at night: a wide of three people at a table, matched singles of Jake and Cohn, a kick under the table, then the two men walking up a lamplit boulevard to a corner kiosk, each drawing labelled with its shot and setup.](https://cdn.moda.app/voyager/blog/script-scene-to-storyboard-animatic/film-7a4b4ca799c736eb.mp4)

23 shots, 86 drawings, about one drawing a second with more on the kick and the walk, cut to the book's dialogue with generated voices, foley, ambience and score. Sound on.

Dialogue from the novel's café scene with Jake, Robert Cohn and Frances. Jake mentions a girl he knows in Strasbourg, Frances kicks him under the table, and Jake suggests Bruges or the Ardennes instead. Jake and Cohn walk up the boulevard to the corner kiosk, Jake calls Cohn back when he forgets his paper, and Cohn walks back to the café. Under the dialogue: café murmur, glasses, footsteps on the boulevard and a quiet string underscore.

**Approve a camera plan before anything is drawn, then draw every frame of a setup as an edit of one master.** Our first passes drew each panel on its own, and the camera moved inside shots, heads changed size and a character ended up on the wrong side of a car. Locking the camera per setup and changing only the performance fixed it; the director's note on the next cut was "way better."

Treat everything below as defaults: it is what worked for one director on one book, written as a starting point. If you already have a workflow, such as your own shot numbering, approval points, board style or sound tools, keep it and take only the parts that help.

## What you need

* **Skills:** `make-image`, `voiceover`, `make-sound` and `make-music`.
* **Plan:** Starter or higher, for the image, voice, sound and music models.
* **Models:** [Nano Banana 2](https://voyager.so/resources/models/nano-banana-2) for text-to-image and reference edits, a described-voice model for the dialogue, and sound effect and music models for the rest.
* **Tools:** FFmpeg for the labels, the line cleanup and the mix, and local transcription for checking every line.
* **Made in Voyager:** this example is a real Voyager run, from the character cards to the mixed cut, with every rule below written from a director's notes on it.

## Read the source and draw characters from it

Pull the scene's exact text first. By default, dialogue comes verbatim from the source, trimmed rather than paraphrased, unless you are adapting it. Our text came from [Project Gutenberg](https://www.gutenberg.org/ebooks/67138), where the novel is public domain in the US.

Then search the source for how each character looks and quote it into their card. Our first Cohn was a second lean, dark-haired man in a suit, and the director couldn't tell him from Jake. The book gives him a nose "permanently flattened" from boxing at Princeton and spectacles, so the redrawn card has round wire rims, a broken nose and rumpled tweed. Where the source is thin, invent contrast: hair shape, eyewear, build, wardrobe.

Watch the words on the card, because they drive every drawing. Jake's first card said "tired deep-set eyes," and every panel he appeared in looked exhausted. Describe the default state you want on screen.

```text
Rough Hollywood storyboard character reference: loose black pencil
and charcoal on white paper, quick grey hatching, strict grayscale,
plain white background. [Source-accurate description: age, build,
face, nose, hair, eyewear, wardrobe, quoted from the text where
possible.] One head-and-shoulders portrait in three-quarter view and
one full body standing. No text.
```

Show the cards and get approval before any scene is drawn. If a redesign comes later, make a new card version and apply it from the next scene, rather than silently redrawing approved ones.

## Cast voices from descriptions and measure them

Build each voice from a written description and audition two or three, rather than starting from a preset. Our first male lead came from a preset that turned out to be a female voice, and an earlier set of voice references came back in Chinese because the preset was bilingual.

```text
English speech only. Clean studio recording. An adult MAN, [age,
accent, register, temperament]. Speak exactly: [line]
```

Use the chosen audition as the reference for every line, and check each take three ways: transcribe it and compare it with the script, confirm the detected language, and measure its pitch. Men should sit around 80 to 130 Hz and women around 170 to 230 Hz; we measured with Praat, because simpler pitch trackers were unreliable on short clips. Retake anything that drifts.

Listen to the delivery too. "Jake's voice is a little too sad" led to lighter auditions, and the lead stayed in the new register from then on. Don't swap a voice the director didn't ask to change.

## Write the shot list and camera plan, then stop for approval

Write the shot list with setup, seconds, drawings, action, speaker, the verbatim line and the state at the start of the shot: what each person wears and holds, and what is on the table. Then draw a camera plan: an overhead per location with the 180-degree line, every camera where a crew could stand, its lens, what sits behind the subject and one eyeline per setup, such as "Cohn looks screen-LEFT toward Jake." Present both on one page with the cost estimate, and wait for approval before drawing. If someone would rather you run straight through, as our director did on the first scenes, say what you would normally stop for and carry on.

![Two overhead camera plans. Inside: Jake and Cohn on either side of a round table with Frances between them, a red 180-degree line through the table, setups A to E marked with lens. Outside: the walk from the café to the corner kiosk, with tracking, two-shot and wide setups F to K.](https://cdn.moda.app/voyager/blog/script-scene-to-storyboard-animatic/camera-plan-2344fbdfd24003d1.webp)

The approved camera plan for the café scene. Every interior camera sits on the room side of the line, so Jake stays screen-left and Cohn screen-right; outside, the walking singles look back up the boulevard, so the kiosk is ahead of them, not behind.

```text
Write the shot list for [scene] from the source text: one row per
shot with setup letter, seconds, number of drawings, the action, the
speaker, the line verbatim, and the state at the start of the shot:
what each person wears and holds and what is on the table. Then make
a camera plan page with an overhead diagram per location: the
180-degree line, every camera position a crew could occupy, lens,
camera height, what each setup sees behind the subject, and one fixed
eyeline per setup (who looks screen-LEFT or screen-RIGHT, at whom). Stage the blocking from
the source; never skip a physical beat. Show me the page and the
cost estimate and wait for my approval before drawing anything.
```

Stage it the way the source does. In our café scene the two men part before the corner, then one calls the other back; the first cut had them already at the kiosk, which the director found confusing. Geography has to make sense too: if they walk toward the newsstand, it is ahead of them, not behind.

## Write a continuity bible for the masters

Write one paragraph that fixes what must not change across the scene: the time of day and where the light comes from, the room's furniture and walls, who sits or walks on which side of the frame, what each person wears in each location, which landmarks stay on which side, and what never appears. Use it to write each setup's description in the camera plan, and give each master the parts of it that its camera actually sees. A test run on another scene with only a one-line setting gave us singles lit flatly next to a two-shot in hard sun, and luggage behind one character in her single but not in the wide.

```text
CONTINUITY BIBLE: a winter NIGHT in 1920s Paris. INTERIOR, the Café
de Versailles: mirrors, brass wall lamps, bentwood chairs, small
marble tables, a dark night window. Jake sits on the LEFT of the
table, Frances in the MIDDLE facing camera, Cohn on the RIGHT. Jake
wears his grey tweed three-piece suit (his dark overcoat hangs on his
chair); Cohn wears round wire-rimmed spectacles and a plain rumpled
tweed jacket; Frances wears her sleeveless evening dress and long
beads. EXTERIOR: a dark boulevard lit by street lamps, bare winter
trees; both men in dark overcoats, Jake on screen-LEFT and Cohn on
screen-RIGHT; the lit newspaper kiosk always stands on the RIGHT-hand
corner. No crowds, no daylight.
```

Keep the whole bible out of the beat prompts. The beats get their continuity from the master image, plus one short line of wardrobe and state for whoever is in frame. Our café beats carried only that line. In a second test run we pasted the full bible into every beat, and continuity got worse: the station building, tracks and hills it described got painted into tight singles that should have shown only a wall behind the subject.

The state line is where props live. Without it they flip between drawings: in the first test run a hat sat on one character's head in one drawing of the two-shot and was gone in the next.

## Draw one master per setup, then stop for approval

Draw a single master frame for every camera setup, with the cast cards of everyone in frame as references. Then check each master before going further: it matches the camera plan's size, lens, side and background; a clean single has no foreground shoulder; reverse pairs match in size and mirror each other's eyelines; and inserts show only hands and props.

By default, stop here too: show the masters on one contact sheet beside the camera plan, and wait for approval before drawing any beats. Every beat copies its master, so a wrong background, side or eyeline in one master repeats in every drawing of that setup; it is far cheaper to fix one master than twenty beats.

Every drawing shares one style string, which keeps 86 frames looking like one artist drew them:

```text
One single full-frame 16:9 rough Hollywood storyboard drawing. Fast
loose black pencil and charcoal marks on WHITE paper, visible
construction strokes, simplified faces, sparse background lines and
quick grey hatching. Strict grayscale, unfinished production sketch,
flat paper, thin black frame border. Unlettered. Use supplied cast
references ONLY for identity and costume, not their finish, detail,
pose or framing.
```

```text
[Style string] CAMERA SETUP [letter], reused for every drawing of
this setup: [size, lens, height, who is where on screen, and only what
this camera sees behind them, taken from the bible]. [Light direction
from the bible.] [The setup's eyeline from the plan.]
STATE: [what they wear and hold, what is on the table]. MOMENT:
[first beat]. Nobody looks into the camera lens. Absolutely no text.
```

## Edit every other drawing from its master, one beat each

Every other drawing in a setup is an edit of that setup's master, so the camera, lens and background can't drift. Each one gets its own specific, physical beat. A generic "different pose" gave us the same frame twice: the director's note was "it's the same shot of Cohn at 12A and 12B."

![Four pencil storyboard frames of the same close shot of Jake at a café table, with the same wall sconce and panelling behind him: holding his glass and smiling, holding up two fingers, raising his glass, and raising one finger to make a point.](https://cdn.moda.app/voyager/blog/script-scene-to-storyboard-animatic/locked-beats-45459fb6058acab1.webp)

Setup B, Jake's single at the table, keeps the same camera, lens and background across shots 2, 4 and 7 while only his performance changes.

```text
[Style string] The FIRST reference image is the LOCKED CAMERA FRAME
for this setup. Keep EXACTLY the same camera position, lens, framing
size, composition, background and lighting, and the same people in
the same places and costumes. Change ONLY the performance, clearly
and visibly different from the reference: [specific beat, e.g. leans
back from the table, glass raised, half-smile]. [The setup's eyeline
from the plan, word for word, e.g. Jake looks screen-RIGHT across the
table, not at the camera.] [One line of wardrobe and state for
whoever is in frame.] Only [name] in frame. Rough pencil storyboard
style. Absolutely no text.
```

Pass the master as the first reference and one cast card per person in that frame; the second test run sent seven images with every beat (the master plus two cards for each of three characters), and the master lost its grip. A beat changes the performance only: when someone enters, leaves or moves to a new place, draw that moment as the master of its own setup instead of an edit. Put the beat first and make it physical, since text-only edits slip back toward the master. Paste the setup's eyeline from the camera plan into every drawing of that setup, word for word, rather than improvising one per drawing: ours drifted into the lens until the director pointed out "they should not be looking at the camera," and 21 drawings needed fixing. In the test run, close-ups of a character seated on the right came back looking screen-right, across the line, even with the right direction in the prompt, which is why the inspection below checks every face. For characters walking away, say "seen from behind, his back to camera, face not visible," or they turn to face the lens.

## Let density follow the action

Aim for about one drawing a second overall, then spend them where the scene moves. Our first pass at one a second felt fast; at half that, the director asked for more frames to make it feel alive, then more again "where there's more action." The version that worked varied: four to six drawings in two seconds on a kick, a door, a pour or a hand-off, timed unevenly so the impact holds and the swing flashes past, and fewer on quiet holds and stares. Dialogue sits around one a second, with in-betweens on the big gestures.

## Escalate coverage with the tension

Open on wides and medium close-ups, move to close-ups as the conflict tightens, and land a tight close-up on the key line. Release to a wide when someone breaks it off. Long runs of identical medium close-ups drew "we need more close ups," and a later scene overcorrected into "too much close ups in the bar scene now." What held up was two to four close-ups per section, on the peak lines only.

Make a close-up by redrawing the setup's master tighter, so the background, light and eyeline stay put. A close-up used more than once is a setup of its own: put it in the camera plan, draw its master this way, and edit its beats from that close-up master like any other setup.

```text
Redraw the FIRST reference image much tighter, as a CLOSE-UP: his
HEAD fills most of the frame height, only the top of his shoulders
visible, background soft. Same man, same expression, costume and
eyeline (looking screen-LEFT, never at the camera). Only him.
```

Avoid reusing an earlier angle for a new story moment. "Why do you reuse shot 21? We already saw this angle" meant cutting the shot or planning a new setup.

## Inspect every drawing on contact sheets

Lay each pass out on contact sheets in cut order, labelled with shot and drawing, and zoom into every face. Check each drawing against its setup's eyeline and its shot's state, not just against its neighbours: a drawing looking the wrong way across the line, or holding a prop it shouldn't, fails. Fix by editing the bad drawing itself, and drop a drawing rather than ship a broken one.

```text
Check every drawing on contact sheets in cut order. Camera and
background identical to its master; a visibly different performance
from its neighbours; only the intended people in frame (no strangers,
doubled faces or reflections); eyes off the lens, toward the scene
partner, matching the setup's eyeline from the camera plan; wardrobe
and props matching the shot's state (spectacles on, hat off, one
glass); light from the same side as the bible; natural proportions; the right side of cars and
tables; day and night consistent. Fix each failure by editing that
drawing, e.g. "REMOVE every other person entirely, ONLY [name] in
frame" or "Keep EVERYTHING identical; change ONLY [name]'s eyes,
turned screen-RIGHT toward [other], not at the camera lens."
```

Our timed notes were exactly these: a man on the wrong side of the car, a head drawn too large, a stranger in an over-the-shoulder, two glasses in two hands.

## Clean the lines, then build the sound

Cut dialogue only inside measured silence. Our first cleanup trimmed on transcription word times, which run early, and clipped the ends of lines. Find the scripted words with transcription, then make every cut inside a stretch of measured silence around them, cap internal pauses at about half a second, and retake any line whose last word runs into the end of the file.

Then add what makes it play: one continuous ambience bed per location, foley spotted to the action, and a restrained score. Keep birds and animal calls out of the beds; a looping bird call in ours read as "a wailing sound." Describe the score rather than naming a composer or director, since prompts with named artists were refused.

```text
Dramatic film score underscore, [tempo], [key], for [dramatic
situation]. Start with just a low sustained cello drone, then bring
in slow, uneasy swelling string chords and a lone distant piano note.
Restrained, brooding, vast and intimate; no melody hooks, no
percussion grooves, lots of space for dialogue. Instrumental, quiet
throughout.
```

Mix to numbers, not by ear. We set lines at -20 LUFS with a limiter, ambience at -40, foley between -30 and -40 and music at -30, ducked under the dialogue with a sidechain compressor, then delivered at -16 LUFS integrated and -1 dBTP. The first music bed was too loud; measured levels fixed it.

## Review it in rounds

Render the cut, show it, and take notes against timecodes; then fix only the drawings and sounds each note names, and render again. Expect the first full render to be a first pass. The café cut above is the eleventh render of that scene: the notes in between moved it from "there's a lot wrong continuity wise" through a restaged walk to the kiosk, 21 eyeline fixes and a cut reused angle, to "omg way better." Ask before changing anything a note didn't mention, such as a voice the director already approved.

## Label it like a real shot list

Burn a small label into the top-left of each drawing: `SH 010 · 8A`, with the setup description beneath. Shots go up in tens, the setup is the scene number plus a letter in order of appearance, and a returning setup keeps its letter. We stopped lettering each drawing, because 10A and 10B read as two different shots, and our director preferred this small label to a full technical slate across the frame. If your production has its own numbering or slate, use that instead. The café cut above predates this numbering note and still reads `SHOT 02` with the setup letter beneath; the scenes after it used `SH 010 · 7A`.

Start with one short scene of two or three people, and **lock the camera plan before the first pencil line.**


## Voyager is a desktop AI agent for creative work

Describe what you want to make; Voyager plans it, uses your apps and tools, and keeps everything editable.

[Sign up](/sign-up)