Wan Video Generation - Overview

Doc version：v1.0.0 | Last updated：2026-07-22
This platform fully supports the official Tongyi Wanxiang (Wan) video generation API. Requests and responses are transparently proxied; parameter semantics are identical to the official API.

Overview

Wan Video Generation supports the following task types:

Task Type	Endpoint	Description
Text-to-Video	`POST /ali/api/v1/services/aigc/video-generation/video-synthesis`	Generate video from a text prompt; wan2.7/wan2.6 support automatic audio and multi-shot narratives.
Image-to-Video	`POST /ali/api/v1/services/aigc/video-generation/video-synthesis`	Generate video using an image as the first frame; supports first-last frame transitions and audio-driven generation (wan2.7).
Reference-to-Video	`POST /ali/api/v1/services/aigc/video-generation/video-synthesis`	Generate video based on subject appearances from reference images/videos; reference them in the prompt with syntax such as `[Image 1]`.
Video Editing	`POST /ali/api/v1/services/aigc/video-generation/video-synthesis`	Edit an existing video by instruction (recoloring, scene replacement, adding/removing elements, etc.), wan2.7-videoedit.
VACE Video Editing	`POST /ali/api/v1/services/aigc/video-generation/video-synthesis`	Five capabilities in one: multi-image reference, video repainting, local editing, video extension, and canvas expansion (wanx2.1-vace-plus).
First-Last-Frame-to-Video	`POST /ali/api/v1/services/aigc/image2video/video-synthesis`	Input a first frame and a last frame to generate a transitional video between them (kf2v series).
Motion Transfer / Character Swap	`POST /ali/api/v1/services/aigc/image2video/video-synthesis`	Make the person in an image imitate the motion in a video (animate-move), or replace the person in a video with the person in an image (animate-mix).
Digital Human Video	`POST /ali/api/v1/services/aigc/image2video/video-synthesis`	Generate a talking/singing digital human video from an image + audio (wan2.2-s2v).
Face Detection	`POST /ali/api/v1/services/aigc/image2video/face-detect`	Pre-check for digital human tasks — validates face usability in the image (synchronous endpoint, returns immediately).
Query Task	`GET /ali/api/v1/tasks/{task_id}`	Poll task status and results

Except for Face Detection, which is a synchronous endpoint, all others are asynchronous tasks: creating a task returns output.task_id, and you must poll the query endpoint for status. On success, the response contains a video download URL (valid for 24 hours) — download or transfer promptly.

Standard Workflow

Create Task (POST https://platform.dataeyes.ai/ali/api/v1/services/aigc/video-generation/video-synthesis)
        │  Returns output.task_id
        ▼
Query Status (GET https://platform.dataeyes.ai/ali/api/v1/tasks/{task_id})   ← Poll until SUCCEEDED
        │  Returns output.video_url
        ▼
Download Video

Endpoint Address

Path Rule：Append the channel prefix /ali after the platform domain, then add the official API path.
Official: https://dashscope.aliyuncs.com/api/v1/services/aigc/video-generation/video-synthesis
→ Platform: https://platform.dataeyes.ai/ali/api/v1/services/aigc/video-generation/video-synthesis

Authentication

All requests use a Bearer Token in the HTTP Authorization header:

Authorization: Bearer {API_KEY}

{API_KEY} is the API key created in your platform dashboard.

Request Headers

Header	Required	Description
`Authorization`	Yes	`Bearer {API_KEY}`
`Content-Type`	Yes	Fixed as `application/json`

The X-DashScope-Async: enable header required by the official API is added automatically by the platform — you do not need to include it; including it is also compatible.

Model Overview

Model	Capability	Submit Endpoint	Key Specs
`wan2.7-t2v`, `wan2.7-t2v-2026-06-12`	Text-to-Video	video-generation	720P/1080P, 2–15 s, with audio, multi-shot
`wan2.7-i2v`	Image-to-Video (first frame / first-last frame / audio-driven)	video-generation	720P/1080P, 2–15 s, with audio
`wan2.7-r2v`	Reference-to-Video	video-generation	720P/1080P, 2–15 s
`wan2.7-videoedit`	Video Editing	video-generation	Instruction-based editing; duration can inherit from the input video
`wan2.6-t2v`	Text-to-Video	video-generation	720P/1080P (size), 2–15 s
`wan2.6-i2v`, `wan2.6-i2v-flash`	Image-to-Video	video-generation	480P–1080P (model-dependent); flash is the cost-effective variant
`wan2.6-r2v`	Reference-to-Video	video-generation	720P/1080P
`wan2.5-t2v-preview`, `wan2.2-t2v-plus`, `wanx2.1-t2v-turbo`, `wanx2.1-t2v-plus`	Text-to-Video (earlier versions)	video-generation	Resolution tiers vary by model; fixed or 5/10 s
`wanx2.1-vace-plus`	VACE Video Editing	video-generation	Fixed 720P, 5 s
`wan2.2-kf2v-flash`, `wanx2.1-kf2v-plus`	First-Last-Frame-to-Video	image2video	flash: 480P/720P/1080P; plus: 720P
`wan2.2-animate-move`, `wan2.2-animate-mix`	Motion Transfer / Character Swap	image2video	Two tiers: wan-std / wan-pro
`wan2.2-s2v`	Digital Human Video	image2video	480P/720P, duration determined by the audio
`wan2.2-s2v-detect`	Face Detection (synchronous)	face-detect	Returns detection results immediately

The actual available models are subject to the model list in the platform dashboard.

Task Status Values

Status	Meaning
`PENDING`	Task queued
`RUNNING`	Task processing
`SUCCEEDED`	Task succeeded
`FAILED`	Task failed
`CANCELED`	Task canceled
`UNKNOWN`	Task does not exist or status unknown (task_id does not exist or has exceeded its 24-hour validity)

Status flow: PENDING → RUNNING → SUCCEEDED / FAILED.

Response Format

All asynchronous endpoints use a unified response pattern:

Create Task Response — returns task_id for polling:

{
  "output": {
    "task_status": "PENDING",
    "task_id": "0385dc79-5ff8-4d82-bcb6-xxxxxx"
  },
  "request_id": "4909100c-7b5a-9f92-bfe5-xxxxxx"
}

Query Task Response — returns task status, video URL, and usage:

{
  "request_id": "caa62a12-xxxx",
  "output": {
    "task_id": "eff1443c-xxxx",
    "task_status": "SUCCEEDED",
    "video_url": "https://dashscope-result-sh.oss-accelerate.aliyuncs.com/xxx.mp4?Expires=xxx"
  },
  "usage": {
    "duration": 5,
    "video_count": 1,
    "SR": 1080
  }
}

The generated video and the task_id are both valid for 24 hours — download or transfer promptly.
Generation typically takes 1–5 minutes (depending on model and duration); polling the query endpoint at roughly 15-second intervals is recommended.

01-Text-to-Video

02-Image-to-Video

03-Reference-to-Video

04-Video-Editing

05-First-Last-Frame-to-Video

06-Motion-Transfer-and-Character-Swap

07-Digital-Human-Video

08-VACE-Video-Editing

09-Query-Task