Seedance 2.5 audio reference: how to send an audio clip

Seedance 2.5 accepts an audio clip as a reference, and BytePlus's tutorial says an audio-only reference is supported on 2.5 while 2.0 needs an image or video alongside it. On Sume, add an audio_url entry to input_references on seedance-2.5.

Sume's catalog for seedance-2.5 lists image_url, video_url and audio_url as accepted reference types, per its video docs.

How do I send an audio reference?

Host the file at a public HTTPS URL and refer to it in the prompt. Sume normalizes tags like @audio1 to the casing the model expects.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedance-2.5",
    "prompt": "A drummer plays in a small club, cuts match the beat of @audio1",
    "duration": 10,
    "resolution": "720p",
    "aspect_ratio": "16:9",
    "generate_audio": true,
    "input_references": [
      {"type": "audio_url", "audio_url": {"url": "https://example.com/beat.mp3"}}
    ]
  }'

What audio files does the vendor accept?

The ModelArk tutorial sets these limits.

Does the reference audio become the soundtrack?

Do not assume it. The reference guides the generation; whether it appears unchanged in the output is not documented on either page. Check a short 480p clip first, and use generate_audio to control whether the model produces sound; see AI video with sound.