Text to video

Request

{
  "model": "wan-3.0",
  "mode": "text_to_video",
  "variant": "standard",
  "parameters": {
    "resolution": "720p",
    "duration_seconds": 5,
    "ratio": "16:9",
    "audio": true,
    "prompt_extend": false,
    "watermark": false
  },
  "input": {
    "prompt": "A paper boat crosses a rain puddle, macro lens"
  }
}

input.prompt is required and must contain 1–20,000 characters after trimming. Set the output duration, resolution, ratio, audio, and watermark in parameters.

Prompt extension

prompt_extend defaults to false. Set it to true to ask the model to expand the prompt before generation.

After create

Save the returned generation ID and poll the status endpoint. Do not keep the create connection open while the video renders.