Skip to main content

Video Generation Live

Generate a video from a text prompt, reference image, or source video. The endpoint queues work for async delivery via callback_url.

Request body

Models

common means callback_url, additional_settings, and enable_fallback. Model-specific provider knobs such as seed, negative_prompt, and generate_audio are passed in additional_settings. generate_audio is only supported on audio-capable models (those that list it below); setting generate_audio: true adds an audio track and roughly doubles the per-second cost. The cost shown below is the default audio-off rate.

Example request

Example response

When processing completes, the result is delivered to your callback_url.

Audio on supported models

Audio-capable models (those that list generate_audio in the table above, such as seedance-2.0-spicy) accept generate_audio inside additional_settings. It defaults to off; setting it to true adds an audio track and roughly doubles the per-second cost.