---
name: animate-scene-pairs
description: "Put your AI character through a whole film scene longer than 15 s (a minute, many shots) on MiniMax H3 Max Turbo image-to-video — one generation = two shots (first frame of shot A, last frame of shot B), the end image is a face board (the character's face full-frame on the left, the true last frame on the right, on black) that stops the face drifting, and the board tail is cut off in the edit. Tested on a 71 s, 18-shot scene in 10 generations. Use when the user says whole scene, long scene, a minute of film, face drifts, keep the face consistent across shots. Part of the animate meta-skill."
---

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-scene-pairs · a whole scene in start/end-frame pairs with a face board

> 🏆 **Source quality recommendation:** Seedance 2.5 reference-to-video (`bytedance/seedance-2.5/reference-to-video`, references as @Image1… / @Audio1…, 4–30 s, native sound) is the supplier's preferred quality route. That recommendation is supported by the product-commercial comparison, not a measurement of every character task. The H3 methods below retain their own measured inputs, settings and timings. Use supported external tools; this is not a FlashBoards integration claim.

{film} · {scene} longer than 15 s → pairs of shots on H3 Max Turbo image-to-video at 768P, joined with the film's sound, the person's {character} in every shot. Tested: 71 s, 18 shots → 10 generations.
Evidence: https://flashboards.yaroflasher.com/learn/acting/ai-character-sheet/#whole-scene
A scene of 15 s or less → `animate-film-scene-refs` or `animate-storyboard-turbo`.

## Choose

| Priority | Settings | Time | Result |
| --- | --- | --- | --- |
| ⚡ takes | `minimax/h3-max-turbo/image-to-video` · **768P** · each pair at its length + ~1 s | 81 s generated across ten pairs; timing not recorded | poses anchored at both ends of every pair, the face held by the board |
| 🏆 final | the same ten again at **1080P**, joined, then one Starlight Fast 3 pass at 1080p (`upscale-video-final`) | not recorded here | skin, hair, fabric and set textures redrawn |
| without the board | start + end frame only | — | ✕ the face drifts: rounder, older, other hair, at times another person |

## Steps

1. **Cut the scene at every cut and pair the shots**: one generation = two shots, 5–15 s. A long single shot goes alone (in the test: 12.4 s and 10.7 s).
2. **Keyframes from the film itself**: the first frame of the first shot and the last frame of the second — never from the other side of a cut.
3. **Replace the actor** in keyframes that show the actor — GPT Image 2.5 Sunburst edit with your own {character sheet} as picture 1. Use the replacement template in `animate-film-scene-refs`: picture roles only, with nothing of the actor's clothes, headwear or gear retained. The historical tested version named the clothes; the later blind *Mad Max* test also held 4 of 4. Use your own inputs, not a demo sheet or prompt. **Frames with nobody in them go in as they are** (an empty street, a button, a closed door): replacing them made the model invent a character who was never there.
4. **Build the end image as a face board**, two full frames side by side on black: left — {character}'s face as close as it goes, brows to lips, filling a 16:9 frame (ours 1920×1080), cut from the person's {face board} / sheet; right — the true last frame of the pair. Ours were 3840×1536.
5. **Write the prompt shot by shot** (camera, action in three steps, every turn as a side of the frame, the music in words) and add **one frozen board shot after the scene**, with `cut_at` = the pair's length + 0.2 s. Four changes against a plain start/end prompt:
   - `output` — "… Then a hard cut to the given last image, which stays on screen as a completely still picture until the end of the video. The video begins exactly on the given first frame."
   - `keyframes.last_frame` — "a still reference board of two full frames side by side on black: picture 1 on the left — only a reference for {character}'s face (an extreme close-up of {character}'s face filling the frame, from the brows to the lips): never use its framing, background or light in any shot; picture 2 on the right — the end of shot 2, the true last frame of the scene … It appears only in shot 3, after the scene is over."
   - `keyframes.rules` — every shot is full screen, "the pictures are never shown together", "the screen is never split and the board never appears", the board shot is frozen.
   - `characters.{character}.face` — "exactly the face in picture 1 (on the left) of the last image — the same face, skin, every mark and the same hairline in every shot …" (blind: point to the picture, do not list someone's features).
6. **Generate the pairs** through a supported external tool using Turbo, the first frame as `image_url`, the face board as `end_image_url`, the pair prompt and duration, `resolution: "768P"` and prompt enhance disabled. Duration = the pair's length plus about a second, whole seconds, at least 5 (ours 5–13 s).
7. **Cut off the board**: find the cut, keep the scene before it, fit it to the film's length (the model usually gives the scene a few tenths of a second more). If a generation lingers on its start frame, drop that moment too. Join the pairs and lay the film's sound under them.

## What to expect

- Two shots per generation keep the poses: each shot is anchored by a real frame at one end.
- The board adds about a second of generation per pair and one cut in the edit; it never reached the finished video.
- Very short shots come out a little fast after fitting.
- Keep per-pair data next to the takes: shots, length, duration, board cut (`meta.json`) and the pair prompt.

## Never

- Only a start and an end frame over ten seconds of close-ups and turns — the face drifts.
- A keyframe taken across a cut.
- Replacing a frame that has nobody in it.
- Leaving the board tail in the edit.
- A demo sheet or prompt in place of your own character material.

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

The stairs dance from *Joker* (2019), 71 s, 18 shots, 10 generations, with the demo character Mara Vale. Empty frames kept as they were: the empty alley, the call button, the closed lift door. Without the board the face came out at times as another woman; the face line listed Mara's freckles, the scar through the outer end of her left eyebrow and her mole; the 1080P final redrew brick. Keep your own pair prompts and metadata beside the takes; the private example files are not supplied.