Background music for a 3-minute Short: one $0.125 generation

A background track for a Short costs $0.125 on Sume: one Music Router generation, fixed price, no per-second charge. YouTube's Help page says Shorts can be up to three minutes in square or vertical format (read 2026-10-08), and Sume's music prompt is where you ask for that length, since the API has no duration field.

Match the bed to the cut

Ask for the length you need plus a little, then trim in the edit. The Music 1.0 page says the model makes structured songs up to a few minutes, steered by the prompt.

Prompt for a loud voiceover

Narration sits on top, so write for it: sparse mids, no lead melody competing with speech, and a named moment where it can swell. Add "Instrumental, no vocals" so lyrics do not fight the voice.

curl -X POST https://api.sume.com/v1/music-router/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: shorts-bed-001" \
  -d '{
    "prompt": "Minimal chillhop bed under narration, 92 BPM, F major. Soft Rhodes, side-chained pad, no lead melody. A 70-second track, gentle lift at 0:55. Instrumental, no vocals."
  }'

Putting it under the clip

Take the audio artifact from result.artifacts[] where type is audio. To combine it with narration and picture, use Timeline 1.0: put the voiceover on the audio spine and the music in the optional soundtrack bed, which takes gain_db, loop, fade_out_seconds and duck_db (ducking needs a real spine, not silence).

A music bed plus a $0.10 one-minute timeline render stays under a quarter.