/v1/videos audio_url references: Seedance 2 yes, Omni and Kling no

Whether you can send an audio_url reference depends on the model. On Sume, seedance-2 lists image_url, video_url and audio_url in supported_input_references. gemini-omni-flash-1.1 takes image and video references but no audio. kling-3 takes no reference_*_urls at all. A reference type outside the list returns a 400.

Who accepts what

The Video Router catalog sets this per model. The seedance-2 example row in the Sume docs lists all three reference types.

How to send one

Audio goes in input_references with type audio_url, next to image and video entries. Having any input_references makes it reference-to-video, unless you also send frame_images, in which case frame_images wins and the request is image-to-video.

{
  "model": "seedance-2",
  "prompt": "The singer performs on a rooftop at dusk",
  "input_references": [
    { "type": "image_url", "image_url": { "url": "https://example.com/singer.png" } },
    { "type": "audio_url", "audio_url": { "url": "https://example.com/track.mp3" } }
  ],
  "resolution": "720p"
}

Check the live row

Always read the live row before you build the body: GET /v1/videos/models returns supported_input_references for each model, and the Video Router doc notes that Omni takes no reference_audio_urls.