AI-Driven Character Schedule Planning & Video Performance Generation
An AI-driven system that implements a complete workflow from character creation → schedule planning → event planning → director script → video performance generation. Supports interactive storytelling with branching narratives, multiple image/video generation models, and both CLI and Web interaction modes.
Event Script Example:
{
"time_slot": "07:00-09:00",
"event_name": "Morning Coffee Visit",
"summary": "Alex drops by Luna's apartment with coffee while she stretches, sharing a quiet moment before his work.",
"image_prompt": "Medium shot of Luna and Alex in Luna's small, art-filled apartment...",
"sora_prompt": "Shot 1: Medium shot. Luna sits on floor stretching... [Cut to] Shot 2: Alex handing coffee...",
"event_type": "N",
"involved_characters": ["Luna", "Alex"]
}
Story Performances:
| Time Slot | 03:00-05:00 Peaceful Exhaustion |
07:00-09:00 Morning Coffee Visit |
09:00-11:00 Solo Canvas Study |
| First Frame | ![]() |
![]() |
![]() |
| Full Video | ![]() |
![]() |
![]() |
Interactive WebUI:
![]() |
![]() |
- 🤖 AI Schedule Planning — Gemini 2.5 Pro driven, with smart energy & mood system
- 🎭 Interactive Event System — R-level (simple choices) & SR-level (multi-stage branching narratives)
- 🎬 Director Script Generation — Cinema-grade shot-by-shot output with dialogue & camera design
- 🎥 Multi-Model Video Generation — nano_banana / seedream (image) + sora2 / kling (video)
- 📱 Mobile-First WebUI — 9:16 portrait mode, 12 iPhone size support, floating transparent UI
- 🔄 Real-Time Attributes — Energy (0-100), Mood, Intimacy (L1-L5 Soulmate)
- 🔁 Auto Retry — Built-in timeout recovery for generation tasks
conda create --name zoo_agent python=3.10 && conda activate zoo_agent
pip install -r requirements.txtEdit config.ini — set your API keys and endpoints:
[api]
api_key = YOUR_API_KEY_HERE
base_url = YOUR_API_BASE_URL
model = gemini-2.5-pro
[image_models.nano_banana]
key = YOUR_NANO_BANANA_KEY
[video_models.sora2]
key = YOUR_SORA2_KEY
[video_models.kling]
key = YOUR_KLING_KEY📖 Full configuration reference:
docs/CONFIGURATION.md
# One-click full pipeline
./run_pipeline.sh luna_001 2026-01-26 --template luna
# Or step by step
python main.py run luna_001 --template luna # Schedule + Director
python generate_performance.py -c luna_001 -t 2026-01-26 # Video generation
python web_interactive_demo.py # Launch WebUIExtract to data/, then run python web_interactive_demo.py to try immediately.
├── src/
│ ├── core/ # Agent modules (scheduler, director, event planner)
│ ├── models/ # Data models
│ ├── storage/ # Config & context management
│ └── video/ # Video generation (unified API, scene processor)
├── data/ # Characters, schedules, events, performances
├── assets/templates/ # 13 preset character templates
├── templates/ # HTML templates (editor + user mode)
├── main.py # Complete workflow
├── generate_performance.py
├── interactive_cli.py
├── web_interactive_demo.py
├── run_pipeline.sh
├── config.ini
└── requirements.txt
| Template | Character | MBTI | Style |
|---|---|---|---|
luna |
Luna | INFP | Artistic / Dreamy — Aspiring artist |
alex |
Alex | ENTJ | Leader / Driven — Tech startup founder |
maya |
Maya | ESFP | Free Spirit — Street musician |
daniel |
Daniel | ISFJ | Quiet / Observer — Bookstore owner |
| Image Model | Provider | Video Model | Provider |
|---|---|---|---|
| nano_banana | Wuyin Tech | sora2 | Wuyin Tech |
| seedream | Volcano Engine | kling | Kling AI |
Supports 4 combinations, concurrent processing, auto timeout retry.
📖 Full CLI documentation:
docs/COMMANDS.md
| Script | Description |
|---|---|
./run_pipeline.sh <id> <date> |
One-click full workflow |
python main.py run <id> |
Schedule + director generation |
python generate_performance.py -c <id> -t <date> |
Video generation |
python web_interactive_demo.py |
Launch WebUI |
python interactive_cli.py <id> <date> --gui |
CLI interactive mode |
./start-tunnel.sh |
Cloudflare Tunnel deployment |
This project is licensed under the MIT License.
AI Character Content Generation System







