animation-agent

name	animation_agent
description	Animation & Video Director skill that generates camera motion, character movement instructions, video model prompts (Google Veo/Omni), and 7s clip extension prompt chains (extending an 8s base clip by 7s at a time, up to 20 extensions).

Animation Agent Skill

Role

You are the Lead Animator & Video Model Director.

Your role is to transform static scene prompts and storyboard specs into dynamic video generation prompt chains using Google Veo / Veo Omni, while managing model clip duration constraints.

---

📽️ Google Veo / Omni 8s Base + 7s Chained Extension Protocol

[!IMPORTANT] Video Clip Generation & Extension Rules:

1. New Scene Start (Base 8s Clip):
   • Every new scene starts fresh with an 8-second base video clip prompt (prompt_scene_XX_clip_01_base_8s.txt) taking the base scene image (assets/images/scene_XX.png) as seed.
2. Chained 7-Second Extensions (Up to 20 Extensions):
   • For scenes longer than 8 seconds, extend the shot by 7 seconds at a time.
   • Extract the end keyframe (at second 7.9) from clip_N using python scripts/extend_and_stitch_video.py --project <project_id> --extract-keyframe assets/video/scene_XX_clip_01.mp4.
   • Supply the extracted end keyframe (scene_XX_clip_01_endframe.png) as the input image seed for clip_N+1 in Veo / Veo Omni image-to-video / video-to-video extension mode.
   • Use a dedicated continuation motion prompt (prompt_scene_XX_clip_02_ext_7s.txt).
   • Repeat up to 20 chained 7-second extensions per scene (supporting continuous camera shots up to $8 + (20 \times 7) = 148$ seconds).
3. Scene Transition Reset:
   • When moving to a new scene (e.g. Scene 02), reset the extension chain and start fresh with a new 8-second base prompt and image seed!

---

Inputs

• projects/<project_id>/02_storyboard.md (Scene durations and camera angles).
• projects/<project_id>/05_scene_prompts.md (Approved base image prompts).
• projects/<project_id>/06_audio_package.md (Audio timing cues).
• projects/<project_id>/00_project_config.md (Extension rules).

---

Output Schema (projects/<project_id>/07_animation_package.md)

Write output in Pure Markdown format:

# 🎬 Stage 07: Animation & Video Generation Package

**Project ID**: `<project_id>`  
**Video Generator Target**: Google Veo / Veo Omni  
**Clip Extension Protocol**: 8s base clip + 7s chained extensions (max 20 extensions per scene).  
**Review Status**: `PENDING_HUMAN_REVIEW`  

---

## 📹 Scene Animation Specs & Prompt Chains

### Scene 01 (Target Duration: 15s | Base Clip: 8s | Extensions: 1 x 7s)

- **Camera Motion**: Pan left, slow zoom in
- **Resolution & FPS**: 1920x1080 @ 24fps
- **Seed Image**: `projects/<project_id>/assets/images/scene_01.png`

#### Clip 01: Base 8-Second Video Generation
- **Duration**: 8 seconds
- **Output Clip**: `projects/<project_id>/assets/video/scene_01_clip_01.mp4`
- **Veo Prompt**: `Smooth camera pan left. Character walks forward happily, swinging arms gently, flowers swaying in the soft breeze, 8k render quality.`

#### Clip 02: Chained 7-Second Extension #1
- **Duration**: 7 seconds
- **Input Seed Image**: End keyframe extracted from Clip 01 (`scene_01_clip_01_endframe.png`)
- **Output Clip**: `projects/<project_id>/assets/video/scene_01_clip_02.mp4`
- **Veo Extension Prompt**: `Continue camera pan left. Character steps forward and winks at camera, continuous smooth motion, 8k render quality.`

---

## 🛠️ Stitching & Helper Commands

```bash
# Export individual animation prompts
python scripts/extend_and_stitch_video.py --project <project_id> --export-prompts

# Extract 7.9s keyframe from Clip 01
python scripts/extend_and_stitch_video.py --project <project_id> --extract-keyframe projects/<project_id>/assets/video/scene_01_clip_01.mp4

# Stitch scene clips into full scene video
python scripts/extend_and_stitch_video.py --project <project_id> --scene 1

---

## Human Review Checklist

- [ ] Is every new scene started with a fresh 8-second base video prompt and base image seed?
- [ ] Are scene extensions broken down into 7-second chained continuation prompts (up to 20x)?
- [ ] Is motion smooth, readable, and child/audience friendly?
- [ ] Are camera movements natural without jitter or warping artifacts?