Skip to main content
Our video endpoints let you generate videos from text prompts (optionally with a reference asset), poll for completion, retrieve job metadata, download the binary content, and delete jobs.

Video generation

Here is an example of creating a video generation job: Parameters:
  • model: The model ID of the model you want to use
  • prompt: Text description of the video to generate
  • input_reference: Optional reference video or asset (URL, data URI, file path, or base64)
  • seconds: Length of the video (default: 4)
  • size: Resolution (default: “1280x720”)
  • fps: Frames per second (default: 16)
  • num_inference_steps: Diffusion steps (default: 25)
  • guidance_scale: Guidance scale (default: 5.0)

Retrieve video job details

Use retrieve to check the status and metadata of a video job:

Download generated video content

Once the job is completed, download the binary content (optionally choose a variant like thumbnail or spritesheet):

Delete a video job

Delete a completed (or failed) job if you no longer need it:

Create and poll until complete

Use the helper to create a job and poll until it reaches a terminal state:
For more details, see the API reference for videos.

Create, poll, and auto-download

This flow uses create_and_poll and immediately downloads the resulting video when the status is completed.