Skip to main content
POST
Create video generation task
The generic video generation API works asynchronously: you submit a creation request to get a task_id, then poll the query endpoint until the task succeeds and returns the video URL.

Features

  • Access multiple video models through one unified protocol by specifying the platform model identifier in model
  • Supports text-to-video and image-to-video (first frame / last frame / reference image / reference video / reference audio)
  • Supports negative prompts, resolution, aspect ratio, duration, watermark, and random seed
  • Media inputs accept public URLs and base64 data URIs
  • Supports callback_url result callbacks (only effective for the seedance channel; when empty, the gateway injects the default callback)

Authentication

Include the Authorization header in the format Bearer YOUR_API_KEY.

Supported video models

Available video models in the Model Square, please refer to Model List.

Quick example

A successful creation returns the task ID:

Next steps

Headers

Authorization
string
required

API Key token (Bearer sk-xxx)

Content-Type
string

Body

application/json
model
string
required

Platform model identifier

negative_prompt
string

Negative prompt

generate_audio
boolean

Whether to generate audio; not supported by some models

callback_url
string

Result callback URL (only effective for the seedance channel; when empty, the gateway injects the default callback)

prompt
string

Text prompt

medias
object[]

Reference media input list (first frame / last frame / reference image / reference video / reference audio; URL or base64 data URI)

duration
integer

Generation duration (seconds)

resolution
string

Resolution tier, e.g. 480p/720p/1080p/4k (subject to model support)

ratio
string

Aspect ratio, e.g. 16:9/9:16/1:1/adaptive

watermark
boolean

Whether to include a watermark

seed
integer<int64>

Random seed

Response

200 - application/json

Generic video generation task creation response

task_id
string

Task ID

status
string

Task status: queued / running