Avatar video soundtrack: send prompt or audio_url, volume 0.05-0.4

The Avatar Video create request has an optional package.soundtrack object, and exactly one of prompt or audio_url is required inside it (not both, no mode field). prompt (1 to 5000 characters) generates a Music 1.0 bed and muxes it under the talking head; audio_url is a public HTTPS file you already have, mirrored into Sume media and muxed. volume is a linear level from 0.05 to 0.4, default 0.15. All of this is from the Avatar Video request schema in Sume's OpenAPI file; the music prompt follows the Music 1.0 docs.

What is the difference between prompt and audio_url?

The schema says the prompt path reserves a Music 1.0 estimate, while the audio_url path mirrors your file and muxes only, with no Music reservation and a mux cost of $0. Localhost and private URLs are rejected for audio_url, and a durable public URL is preferred. Create and preview check the URL shape; full accessibility is checked at generate-video, where a dead URL is a soft fail.

Which should I use?

Use audio_url when you have licensed music already, or when you want the same bed across many videos. Use prompt when you want a new bed per video; write it like any music brief and end with "Instrumental, no vocals" so the bed does not compete with the speaker. Remember that Avatar 1.0 itself is English-only, so a soundtrack does not change the language limit.

What does a soundtrack block look like?

The block sits inside package in the Avatar Video create body; other required package fields are unchanged and are not shown.

{
  "package": {
    "soundtrack": {
      "prompt": "Warm acoustic bed, 92 BPM, G major. Nylon guitar and soft shaker, a gentle lift at 0:15. Instrumental, no vocals.",
      "volume": 0.12
    }
  }
}

What if I want more control than a single volume?

Render the avatar video without a package soundtrack and mix in Timeline, where the bed has gain_db, fade_out_seconds and duck_db. The package field is the short path; Timeline is the control path.