StoryVisAI
How it’s made

The anatomy of a 0:42 film

Video production makes six decisions, in the same order, every time — whether a crew of twelve makes them over three weeks or a model makes them in a minute. Here is one sentence going through all six.

Scroll to follow it through

0:00Stage 1 / 6

One sentence goes in

No storyboard, no shot list, no brief document. A single line of intent, plus the two or three constraints that actually matter: how long, what shape, what mood.

You now have: A prompt and three constraints.

Your idea

A lighthouse keeper finds a message in a bottle written in his own handwriting, dated a week in the future.

16:9Cinematic noir~70 seconds

input.txt — 1 sentence, 27 words

0:04Stage 2 / 6

It becomes beats, not paragraphs

The sentence is expanded into a structure with timecodes — because a video is a sequence of decisions about when, and prose hides that. Every beat is editable before anything renders.

You now have: A timecoded script you can rewrite.

Script & beats

  • 0:00Wide. The lamp turning against a bruised sky.
  • 0:14Elias on the rocks. The bottle catches the light.
  • 0:31Close on the note. The handwriting resolves.
  • 0:49The switch. His hand, not moving.
  • 1:04Dusk finishes. Black on an unlit lamp.

script.json — 5 beats, timecoded

0:11Stage 3 / 6

Each beat is assigned a shot

Framing is chosen per beat — wide to establish, close to land the turn. This is the step that decides whether the piece reads as film or as a slideshow with narration over it.

You now have: Nine framed shots with composition notes.

Storyboard

01EWS
02WS
03MS
04CU
05ECU
06WS

storyboard — 9 shots, 6 shown

0:19Stage 4 / 6

The people stay the same people

A character is locked once — face, build, wardrobe, seed — and reused across every shot that needs them. Continuity is the difference between a film and a set of images that happen to share a subject.

You now have: Locked characters, reusable in the next project.

Cast & continuity

E

Elias

60s · weathered · grey wool coat

seed 4417face lockedwardrobe locked
01
04
07
09

Same face in shot 09 as in shot 01 — that is the part most tools lose.

cast — 1 character, seed locked across 9 shots

0:28Stage 5 / 6

Timing is set by the read, not the grid

The narration is generated first, then the shots are cut to it. Captions come out word-aligned as a by-product, so the accessible version is not a separate job.

You now have: A voice track, captions, and a real edit length.

Voice & timing

0:000:441:12

Don’t light the lamp tonight,” it warned.

Elias — low, weathered-14 LUFS

voice.wav — 1:12, captions word-aligned

0:36Stage 6 / 6

One pass, every shape you need

Vertical, square and widescreen come out of the same render — reframed per ratio rather than letterboxed. The script stays attached, so the next version is an edit and not a restart.

You now have: Finished cuts in three ratios, plus the source script.

Render & deliver

16:9Landing page
9:16Reels / Shorts
1:1Feed
Rendercomplete
MP4 · H.264SRT + VTTScript archived

exports — 3 ratios, MP4 + captions, one render pass

The same six decisions, either way

Nothing in the list below is skipped or automated away. What changes is how long you wait between making a decision and seeing it.

DecisionTraditionalWith StoryVisAI
Script and beatsWriter, 2–5 daysSeconds, then you edit
Shot list and boardStoryboard artist, 3–7 daysGenerated per beat
Casting and continuityCasting, fittings, reshootsLocked once, reused
Voice and timingBooth session, 1–2 daysGenerated, word-aligned
Edit and gradeEditor, 3–10 daysCut to the narration
Versions and ratiosA new edit per formatOne render pass

Questions about the pipeline

Do I have to go through every stage myself?
No. The full sequence runs from a single prompt without any intervention. Each stage is exposed so you can step in where it matters — rewrite a beat, swap a shot, relock a character — and leave the rest alone.
Can I start from the middle instead of a prompt?
Yes. If you already have a script, you can enter at the storyboard stage and skip generation entirely. The same applies to characters and voice — anything you bring is used as-is.
What keeps a character looking the same across shots?
A character is locked once — appearance, wardrobe and generation seed — and that lock is reused by every shot referencing them, including in later projects. Continuity is carried by the character record, not by re-describing them in each prompt.
Why is the voice generated before the edit?
Because the read decides the length. Cutting picture to a finished narration track gives you real shot durations and word-aligned captions as a by-product, instead of stretching or trimming audio to fit a timeline that was guessed first.
6 stages, one prompt

Run the whole sequence on your own idea

Write one sentence and watch it come out the other end — or start from a prompt that already works.