Theme
Omni video models
Omni video models submit async tasks via POST /v1/videos and query status via GET /v1/videos/{task_id}.
Video generation
Text-to-video, first-frame image-to-video, first–last frame, and multi-reference-image video
View docs →Reference-video generation
Reference-video edit, extend, and regenerate
View docs →Supported models
| Model | Description | Supported modes | Reference limits |
|---|---|---|---|
| General fast video generation model | Text-to-video, first-frame image-to-video, first–last frame, multi-reference images | Up to 5 ; reference video not supported | |
| Reference-video generation model | Reference-video edit, extend, regenerate; compatible with some T2V / I2V fields | Reference video max 15MB |
Supported capabilities
| Capability | Key fields | ||
|---|---|---|---|
| Text-to-video | Supported | Supported | , , , |
| First-frame image-to-video | Supported | Supported | |
| First–last frame | Supported | Supported | , |
| Multi-reference images | Supported | Supported | |
| Reference video | Not supported | Supported | / / |
Unified entry point
http
POST /v1/videosAll Omni video tasks use a JSON request body. On success, the response is an OpenAI-style video object; use id or task_id for subsequent queries.
Endpoint summary
| Capability | Method | Path |
|---|---|---|
| Video generation | POST | |
| Reference-video generation | POST | |
| Task query | GET |