Text-to-Video

Create a video by describing its scene, subject, action, and visual style. First choose a text-to-video model in the Playground; model IDs, endpoints, and supported controls vary, so use the selected model's page for the exact request schema.

Write a prompt

Describe the important elements in a clear order:

  1. Subject and setting: who or what appears, and where.
  2. Action: what changes over the clip.
  3. Camera: framing or movement, such as a slow push-in or overhead tracking shot.
  4. Lighting and style: time of day, atmosphere, and visual treatment.

For example: A red paper kite rises above a quiet coastal town at sunrise. The camera follows it upward in a slow tracking shot; warm light, soft sea mist, natural colors.

Avoid combining conflicting directions. Add duration, aspect ratio, resolution, or a negative prompt only when the selected model supports those parameters.

Submit and poll

Use the endpoint and request fields shown on the selected model's Playground page:

  1. Submit a POST request with your API key and prompt.
  2. Save the returned request_id.
  3. Poll GET /api/v1/predictions/{request_id}/result until the task completes or fails.

For the shared response format, see the API Reference. For an overview of text-to-video and image-to-video workflows, see Video Generation.

See also

  • Models — discover current models.
  • Kling — model-specific guide.
  • Webhooks — receive task completion notifications.