ACE-Step on Sume: 240 s max, billed per second not per minute

ACE-Step runs on Sume as ace-step with duration_seconds from 5 to 240 (default 60), and it is the only length-controlled music row billed per second of audio. A 45-second clip costs three quarters of a 60-second one; the other rows round up to whole minutes or charge one price per track.

fal lists "$0.0002 per second of generated audio" and returns a WAV (read 2026-10-14). Sume sells at the fal list price times 1.25 and shows credits only.

Request

Pin the id and the length. The row refuses image_url and duration.

The prompt carries style tags and optional lyrics; the result lists the model-reported lyrics when it returns them.

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: ace-45s-001" \
  -d '{
    "model": "ace-step",
    "prompt": "Synthwave drive, 118 BPM, instrumental",
    "duration_seconds": 45
  }'

Per-second billing next to the others

Per-second billing means you can ask for the exact length. That matters for short stings and for cut-to-length beds, where a per-minute row makes 45 seconds cost the same as 60.

Trade-offs to know

Per-second pricing is cheap, but cost is not the only axis: this is an open-source model, and the sound differs from Lyria and ElevenLabs. Run a short 15-second test first.

Live credits are in GET /v1/music-router/models/ace-step and on the API rate card. The ranges are on the Music Router docs page.