One async Python client for multiple video-generation providers: Seedance, Kling, MiniMax/Hailuo and Wan.
| Provider | Version | API Model ID | Notes |
|---|---|---|---|
| Seedance (ByteDance) | 2.5 | doubao-seedance-2.5 |
30s one-take, native audio |
| Seedance | 2.0 | doubao-seedance-2-0-260128 |
Full 2.0, multimodal input |
| Seedance | 2.0 Mini | doubao-seedance-2-0-mini-260615 |
Budget tier, 480p/720p, 4–15s |
| Seedance | 2.0 Fast | doubao-seedance-2-0-fast-260128 |
Faster 2.0, lower quality |
| Kling (Kuaishou) | 3.0 / 3.0 Omni | kling-v3 / kling-v3-omni |
Flagship, multi-shot, native audio |
| Kling | 2.6 | kling-v2-6 |
First with synced audio + video |
| Kling | 2.5 Turbo | kling-v2-5-turbo |
1080p, cheap for high volume |
| Kling | 2.1 | kling-v2-1-master |
Cinematic control, better semantics |
| MiniMax / Hailuo | H3 | MiniMax-H3 |
4K, 60fps, 30s, native audio |
| MiniMax | 2.3 | MiniMax-Hailuo-2.3 |
1080p, T2V + I2V |
| MiniMax | 2.3 Fast | MiniMax-Hailuo-2.3-Fast |
Faster 2.3, I2V only |
| Wan (Alibaba) | 3.0 | wan3.0 |
All-in-one, 30s, up to 1080p |
| Wan | 2.7 | wan2.7-t2v / wan2.7-i2v / wan2.7-r2v |
Specialized models |
| Wan | 2.6 | wan2.6-t2v / wan2.6-i2v |
Multi-shot, longer clips |
| Wan | 2.5 | wan2.5-t2v-preview |
1080p, audio support |
git clone https://github.com/letorig/video-generator-client
cd video-generator-client
python -m venv .venv
# Windows:
.venv\\Scripts\\activate
# macOS/Linux:
source .venv/bin/activate
pip install -e ".[web]"Copy .env.example to .env and add the credentials for the providers you use.
Output is saved to the output folder.
video-gen --help
video-gen providers
video-gen models wan
video-gen generate wan --prompt "A snowy forest at dawn" --model 2.1 --duration 5
video-gen status wan TASK_ID
video-gen serveWeb server options:
video-gen serve --host 127.0.0.1 --port 8000
video-gen serve --host 0.0.0.0 --port 9000 --reloadThe web interface is bundled directly into the Python package; there is no npm, Node.js or frontend build step. FastAPI serves the HTML and exposes JSON endpoints for provider discovery, generation, task polling and MP4 download.
The UI provides:
- provider/model selection
- text-to-video and image-to-video input
- negative prompt
- duration, aspect ratio and resolution
- live task status polling
- video preview and MP4 download
- in-memory task history for the current server session
unified-video-gen/
├── src/video_gen/
│ ├── client.py
│ ├── cli.py
│ ├── web.py
│ ├── catalog.py
│ ├── static/index.html
│ ├── providers/
│ └── utils/
├── examples/
├── tests/
├── mcp_server/
├── .env.example
├── .gitignore
├── LICENSE
└── pyproject.toml
Provider APIs and model IDs change independently of this SDK. Treat the provider adapters as integration code that may need updates when a provider changes its API.
MIT.