Skip to main content
POST
The duration is fixed at 10 seconds, the resolution is fixed at 720p, and the final video includes a generated audio track. video_length and resolution can be omitted.Reference videos, reference audio, first and last frames, seed, and negative_prompt are not supported, and the audio track cannot be disabled.

Authorizations

string
required
All endpoints require Bearer Token authentication.Get the API Key:Visit the API Key Management page to get your API Key.Add to the request header when using:

Body

string
required
Model name, fixed to muse_video_ext.
object
required
Model parameter object.
string
required
Initiator task ID.User-defined task identifier, required. Only letters, numbers, underscores, and hyphens are supported, with a length of 1-64 characters. This ID can be used later to query task status and results.

Response

string
Initiator task ID, can be used to query results.
string
Initial status, typically "pending".
string
Creation time, ISO format.

Query Task

This model uses asynchronous video generation. After submitting a task, please use out_task_id to query the task status. See also Get Task Status.
string
required
The initiator task ID passed when submitting the task.
After the query succeeds and is completed, the output array in the response will contain the generated video, with elements containing the video field. Generation time varies depending on the input: text-only usually takes about 1 to 2 minutes, while multiple reference images may take several minutes. Please continue polling and do not resubmit.

使用说明

Effects of multi-shot prompts:
  • Reliable cut timing: For timestamps specified in the prompt, the actual cut error is within 1 second.
  • Text, prints, and logos on products remain consistent across shots.
  • Characters may differ from the reference image after a cut. When the same person needs to remain recognizable, it is recommended to use at most 2 cut points and repeat their appearance description in each shot.

Request Examples