Wan video

Call Wan (Tongyi Wanxiang) via the unified async API:

Action	Method	Path
Create	`POST`	`/v1/video/generations`
Get result	`GET`	`/v1/video/generations/{task_id}`

POST|GET /v1/videos is also accepted. Host: https://api.omniall.ai · Auth Authorization: Bearer sk-....

Models

Currently supported (Alibaba Wan):

Supports up to 10 reference images, 5 audios, 5 videos; duration up to 30s. See the Model Square.

Scenario	Key fields
Text-to-video	`prompt` + `duration` + `ratio` + optional `generate_audio`
Multi-image reference	`images` / `reference_images`
Omni multimodal	`reference_images` + `reference_videos` + `reference_audios`
First / last frame	`first_frame` / `last_frame` (or aliases / content roles)

Do not mix images with first/last frames.