Qwen Video

Create Wan and HappyHorse text-to-video, image-to-video, reference, and editing tasks.

Endpoints

POST /v1/services/aigc/video-generation/video-synthesis
POST /v1/videos/generations
Authorization: Bearer $API_KEY
Content-Type: application/json

Both paths are equivalent and accept the same body. The upstream accepts the task immediately; poll via GET /v1/tasks/{taskId}.

Models

Model	Capability	Resolutions	Duration
`qwen/wan3.0-video`	text, image (first/last frame), reference, edit, extend	480P/720P/1080P	2–30s or `-1` smart
`qwen/wan3.0-video-prime`	same as above (high-speed)	480P/720P/1080P	2–30s or `-1` smart
`qwen/wan2.7-t2v`	text to video	720P/1080P	2–15s
`qwen/wan2.7-i2v`	image to video (first / first+last frame)	720P/1080P	2–15s
`qwen/wan2.7-r2v`	reference to video	720P/1080P	2–15s
`qwen/wan2.7-videoedit`	video editing	720P/1080P	2–10s
`qwen/happyhorse-1.1-t2v`	text to video with audio	480P/720P/1080P	3–15s
`qwen/happyhorse-1.1-i2v`	image to video, first frame, with audio	480P/720P/1080P	3–15s
`qwen/happyhorse-1.1-r2v`	reference to video (1–9 reference images)	480P/720P/1080P	3–15s
`qwen/happyhorse-1.0-video-edit`	video editing	720P/1080P	3–15s

Request body

Field	Type	Required	Description
`model`	string	Yes	Complete public model ID.
`input.prompt`	string	Conditional	Text prompt; required when no media is provided.
`input.media`	array	Conditional	Media item list; required when no prompt is provided.
`parameters.resolution`	string	No	Defaults to `1080p`. Accepts `480p`, `720p`, `1080p`.
`parameters.ratio`	string	No	Aspect ratio; `adaptive` is recommended for image-to-video and editing modes. Accepts `21:9`, `16:9`, `9:16`, `1:1`, `adaptive`, and others.
`parameters.duration`	integer	No	Seconds; `-1` lets the model pick the duration automatically. `wan3.0` defaults to `-1`; others default to `5`.
`parameters.audio`	boolean	No	Whether to generate synchronized audio; defaults to `true` for the HappyHorse series.
`callback_url`	string	No	Public terminal callback URL handled by the platform.

Each media item contains a type and url (public URL or base64):

`type`	Description
`first_frame`	First-frame image
`last_frame`	Last-frame image (requires a first frame)
`reference_image`	Reference image for reference-to-video
`reference_video`	Reference / edit / extend video
`reference_audio`	Reference audio
`video`	Source video for editing

Examples

curl "https://api.modelrover.ai/v1/services/aigc/video-generation/video-synthesis" \
-H "Authorization: Bearer $API_KEY" \
-H "Content-Type: application/json" \
-d '{
  "model": "qwen/wan3.0-video",
  "input": {
    "prompt": "Detective noir: a rain-soaked New York street at night, neon signs flickering, a detective in a black trench coat walking briskly"
  },
  "parameters": {
    "resolution": "1080p",
    "ratio": "16:9",
    "duration": -1
  },
  "callback_url": "https://example.com/hooks/video?token=unguessable"
}'

Create response

{
  "output": { "task_id": "9dcf8ab2-c4b7-42be-a516-597223897b99", "task_status": "PENDING" },
  "request_id": "..."
}

output.task_id is rewritten to the platform task ID.

Task response

curl "https://api.modelrover.ai/v1/tasks/9dcf8ab2-c4b7-42be-a516-597223897b99" \
-H "Authorization: Bearer $API_KEY"

{
  "output": {
    "task_id": "9dcf8ab2-c4b7-42be-a516-597223897b99",
    "task_status": "SUCCEEDED",
    "video_url": "https://cdn.qwencloud.com/..."
  }
}

Task statuses are PENDING, RUNNING, SUCCEEDED, and FAILED. Video results are retained for 48 hours—download and transfer them immediately after completion.

Restrictions

For wan3.0-video, first-last-frame mode cannot be combined with reference media; different task-type media roles cannot be mixed.
parameters.duration: -1 is supported by the wan3.0 series only; passing -1 to other models returns an error.
Client routing and paid service-tier selections are rejected.

See Tasks and callbacks for polling, callback delivery, and task lifecycle.