---
name: animate
description: "Router for animating an AI character with MiniMax H3 — a film scene from frames and its sound, a storyboard take on Turbo, a whole minute in shot pairs with a face board, motion from a video, voice from audio, character + place + prop images, props that hold, prompt enhance. Classifies the job and the priority (speed / fidelity / detail), then hands off to the matching animate-* sub-skill with exact settings, prompts and times. Use when the user says animate my character, bring the character to life, reference-to-video, put my character into a film scene, lip-sync, voice, dance like this clip, storyboard to video, MiniMax H3, анимировать персонажа, оживить персонажа, персонаж в сцене фильма, липсинк. Two external-tool routes: the measured H3 method and the supplier's quality recommendation, Seedance 2.5 reference-to-video; broader character-format parity was not measured."
---

These creative references describe source experiments with external tools. Use your own supported account and tools; this download installs no automation. Settings and timings are experimental observations, not current FlashBoards availability.

# animate · meta-skill (router)

Work for the person's own character(s): their photo / sheet / voice from their character card. The example runs below belong to demo characters.

Every recommendation here is measured in source experiments (29.09.2026, MiniMax H3 Max and H3 Max Turbo on fal). The methods are generalized character/film templates; the actual test material is listed in the last section.
Article with all the evidence: https://flashboards.yaroflasher.com/learn/acting/animate-ai-character/
The whole-minute scene (shot pairs + face board): https://flashboards.yaroflasher.com/learn/acting/ai-character-sheet/#whole-scene

The character always comes in as the person's **{character sheet}** (face close-up + full body in one image), made from their {photo}. No sheet yet → make one from their photo first. Never send a demo character's sheet, prompt, frames or voice for a person's job.

## Two external-tool routes

| Route | Model | When |
|---|---|---|
| Measured H3 workflow | MiniMax H3 Max / H3 Max Turbo, 480P | drafts and many takes |
| Supplier quality recommendation | Seedance 2.5 reference-to-video (`bytedance/seedance-2.5/reference-to-video`, `task: reference`), pictures as `image_urls` (@Image1…), audio as `audio_urls` (@Audio1…), 4–30 s, native synced sound, 720p / 1080p | final cuts where detail matters |

The supplier prefers Seedance 2.5 for quality. In the source's 15 s product-commercial comparison at 720p, the listed cut deviations reached about 1 s (versus up to 2 s on H3), liquids and fine detail improved, and the model added synced effects; generation took about 7 min versus 9 s on H3 Max 480P. The measurements in `ad-spot` take precedence over the source summary's smaller bounds. This was not measured on every character format. Use the same shot-by-shot structure in plain text with @Image1 and @Audio1 reference names; these are external tools, not a claim of FlashBoards integration.

## 1 · Classify the job

| Input / goal | Go to |
| --- | --- |
| A film scene up to 15 s: its frames + its sound, one request | `animate-film-scene-refs` |
| The same scene, using the tested faster route — one storyboard image + the sheet + pinned audio | `animate-storyboard-turbo` |
| A whole scene longer than 15 s (a minute, many shots), the face must not drift | `animate-scene-pairs` |
| Borrow motion and camera from a clip (a dance, a walk) | `animate-video-ref` |
| A voice line / lip-sync, or a soundtrack the picture must follow | `animate-audio-ref` |
| No film frames: character + location + prop as separate images | `animate-multi-image` |
| An unusual prop that must keep its shape (glasses with three lenses, a strange tool) | `animate-props` |
| Which `prompt_expansion_mode` to send | `animate-prompt-enhance` (answer: `disabled` for anything scripted) |

## 2 · Pick the reference layout

Choose by speed, detail and fidelity. For a film scene up to 15 s, the source's H3 default is Turbo with the 3×3 storyboard first, your {character sheet} last and film audio pinned, both reference images far off the timeline. That experiment generated all five shots in 9 s, against 24.1 s on H3 Max. Use Max reference-to-video when references have different jobs (motion, voice, place, prop).

| | Engine | Why |
|---| --- | --- |
| ⚡ | H3 Max Turbo image-to-video | 9.0 s generation for 15 s (but the fal queue took 30 s and 2 min 33 s on our two runs) |
| 🏆 | H3 Max reference-to-video · 768P | a slot per reference (9 images, 3 audio, 3 video); keeps the sheet off screen without trimming |

## 3 · How the skills link — the full pipeline

```
{character sheet} of the person's character (see the AI Character Sheet article)
        │ unusual prop? put it on the character first
        ▼
props ───────────── animate-props            (GPT Image 2.5 still → start frame → Turbo)
        ▼
frames of the scene ── replace the actor with {character} in every frame (GPT Image 2.5)
        │                 frames with nobody in them stay as they are
        ▼
pick the engine
   ≤ 15 s, all references in one request ── animate-film-scene-refs  (H3 Max r2v)
   ≤ 15 s, storyboard ─────────────────── animate-storyboard-turbo (Turbo, board + sheet far off the timeline)
   > 15 s, many shots ──────────────────── animate-scene-pairs      (Turbo, 2 shots per generation, face board)
   motion / voice / place + prop ───────── animate-video-ref · animate-audio-ref · animate-multi-image
        ▼
prompt ──────────── JSON shot by shot, enhance disabled ── animate-prompt-enhance
        ▼
edit ────────────── trim the board / sheet / face-board frames, fit to the film, lay the sound
        ▼
one final upscale ── upscale-video-final (upscale meta-skill)
```

## 4 · Global rules (break none)

1. **Replace the real actor in every film frame first** (GPT Image 2.5). H3 accepted a frame with the actor without any refusal and copied the actor's face, hair and shirt. Frames with nobody in them go in as they are — replacing them invents a character who was never there.
2. **Say what the sheet is for**: "only the look of {character} … it never appears on screen as a shot". Without it H3 Max opened the video with the sheet for the first second.
3. **Blind wording**: a reference is named by its role, never re-described — "Image 1 is only the look of {character}: exactly their face, hair and outfit from Image 1, nothing added". The look comes from the picture; no face, hair or clothes of anyone are written into a template. Exception (owner, 9.10.2026): the expression of a reference frame — gaze, brows, mouth — may be named.
4. **On Turbo, first/last images that are references stand far off the timeline** (first at -00:10, last at +01:40, no frozen shots for them). As frozen shots they took 0.38 s and 1.8 s of screen and a whole shot was lost.
5. **`prompt_expansion_mode: "disabled"`** for every scripted scene. Balanced invented a cut at 3.0 s and the video followed the rewrite.
6. **Cuts land early.** Asked 5.29 / 7.38 / 8.63 / 11.30 s, got 3.21 / 6.13 / 8.17 / 9.46 s (0.5–2.1 s early) on Max, almost the same on Turbo. The order holds, the timing does not — plan the edit for it.
7. **An audio reference is played back, not re-performed** (correlation 0.89 for music, 0.96 for a voice) — so a voice line must already be {character}'s voice. A video reference gives motion and camera, never the costume.
8. **A prop is given as a picture on the character, not described in words** — "always", positions and negations all broke it.
9. **Images under 256 × 256 are refused** ("image_too_small"): pad on a canvas, do not scale.
10. Heavy compute only in the cloud (fal); local ffmpeg for trims and comparisons only.

## 5 · Settings and times (measured)


| Job · settings | Time |
|---| --- |
| Film scene, 14.8 s, 5 shots · H3 Max r2v · 9 images + 14.8 s audio · 15 s 768P | 24.1 s generation, 34.7 s request, queue 0 s |
| Same scene · Turbo · board + sheet + audio · 15 s 768P | 9.0 s generation, 15–20 s request, queue 30 s / 2 min 33 s |
| Video reference · sheet + 5 s clip · 5 s 480P | not measured |
| Audio reference · sheet + 6.7 s voice · 7 s 480P | not measured |
| Three images · 5 s 480P · enhance disabled / balanced / quality | 9 / 14 / 23 s submit to file |
| Prop recipe · Turbo i2v from the start frame · 7 s 768P | queue 2 s, run 33.4 s |
| Actor replaced in a frame · GPT Image 2.5 Sunburst · low | not measured |
| Whole minute · 10 Turbo pairs · 768P · 81 s generated | not recorded |

## 6 · External tools

Use your own supported account and tool to upload the images/audio, set the model, explicit dimensions and duration, and inspect the outputs. Endpoints measured here: `minimax/h3-max/reference-to-video`, `minimax/h3-max-turbo/image-to-video`, `openai/gpt-image-2.5/sunburst/edit`. No local runners or installed automation are provided.

## Example run (reference only — never reuse its files, prompts, faces, voices or frames for a person)

- Character: the demo character **Mara Vale** (character sheet, 29.09.2026); source experiments only, not reusable input files.
- Film scene rows of §5 (14.8 s, 5 shots): the final drum solo of *Whiplash* (2014).
- Video reference: the stairs dance from *Joker* (2019); audio reference: narration from *Gone Girl* (2014); whole minute: the *Joker* stairs dance, 71 s, 18 shots; actor-replace prompt first written for *Mad Max* frames.