Video Generation and Processing

Video models can expose text-to-video, image-to-video, reference-based generation, editing, extension, or other documented operations. The exact operation set belongs to each model and can differ within the same manufacturer.

Choose a model and operation

Check the selected model for:

• Supported generation and transformation operations.
• Required and optional input modalities.
• Duration, resolution, aspect ratio, and audio controls.
• Reference-image, reference-video, or reference-audio limits.
• Execution mode, result format, retention, and price.

Only send fields declared for the selected operation. Similar model names do not guarantee identical request schemas.

Submit work

Video generation is generally asynchronous. Submit the operation through the public endpoint documented for the selected protocol and model, then save the platform task ID returned by the gateway.

Some protocol families use operation-specific endpoints, while others use long-running model methods or native task resources. These differences belong in the protocol and model reference rather than in the manufacturer list.

Track the task

Use the task-query endpoint associated with the creation protocol until the task succeeds, fails, is canceled, or expires. Where the protocol exposes the platform's common task lookup, use GET /v1/tasks/{taskId}. Where a native task resource is retained, use its documented query path.

When supported, provide callback_url to receive the terminal snapshot. See Synchronous, Streaming, and Asynchronous Requests for polling, callbacks, and retry behavior.

Results and billing

Result URLs can be temporary even when media has been transferred to platform storage. Save assets you need to keep and review Data and Result Retention. Final task status, usage, and cost appear in request logs.