Stable Audio 2.5 on Sume: a 190 s track costs the same as a 5 s one
Stable Audio 2.5 on Sume's Music Router takes duration_seconds 5 to 190 and bills one flat price per track, so ask for the full length you need.

Stable Audio 2.5 is a Music Router model on Sume with the id stable-audio-2.5. It takes duration_seconds from 5 to 190 (default 60) and charges one flat price per track, so a 190-second track costs the same as a 5-second one. Pick the length the edit needs, not the shortest one you can get away with.
Sume's side is the Music Router docs, read 2026-10-14. Fal, which serves the model, lists it at $0.2 per audio generation on its Stable Audio 2.5 page, read 2026-10-14, and says it makes music and sound effects with commercial use permitted. The model returns WAV on fal.
What the router accepts
Numbers below come from the Music Router docs and the Sume pricing code on main, read 2026-10-14.
| Field | Value |
|---|---|
| Router id | stable-audio-2.5 |
duration_seconds | whole seconds, 5 to 190, default 60 |
| Prompt | up to 5,000 characters |
| Vocals | no: instrumentals and sound beds |
image_url | refused (only Music Auto, Lyria 3.5 and Lyria 3 Pro take one) |
| Price basis | one price per track (per_audio) |
What a track costs
Fal lists $0.20 per generation and Sume sells at list times 1.25, which makes $0.25 per track whatever the length. The live number is in the credits on GET /v1/music-router/models and on the API rate card.
| duration_seconds | Cost | Cost per second |
|---|---|---|
| 5 | $0.25 | $0.0500 |
| 30 | $0.25 | $0.0083 |
| 60 (default) | $0.25 | $0.0042 |
| 190 | $0.25 | $0.0013 |
Ask for 190 s, trim in the edit
Because the price does not move with length, the cheap move is to request the longest bed you might use and cut it in your editor or Sume's Timeline. Compare ACE-Step, which bills per second, and CassetteAI Music, which bills per started minute: on those a long request costs more. Rows that do not list duration_seconds refuse it, as covered in which four models take it.
Put exclusions in the positive prompt. Sume does not support a non-empty negative_prompt.
curl -X POST https://api.sume.com/v1/music-router/generate \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "stable-audio-2.5",
"prompt": "Warm ambient synth bed, 70 BPM, soft pads, no drums, no vocals, slow evolving texture, loopable",
"duration_seconds": 190
}'Then fetch the audio
The call returns a job. Poll GET /v1/jobs/{id}/status, then read GET /v1/jobs/{id}/result and take the artifact whose type is audio. job.request.routed_model names the engine that ran.
Sources
Related posts
More in Models
- AI video API clip length: which Sume models accept it, and the cost
Which Sume video models accept a clip length by API, which refuse it at validation, and what each accepting model costs, one row per clip length.
- AI video clip price on Sume, by model, length and resolution
The price of one AI video clip on every Sume video model, by clip length and resolution, as each of these posts read the catalog.
- Cheapest AI video model on Sume by clip length and resolution
Which Sume video model makes a clip of a given length and resolution for the least, and what the dearest model that can make it costs.
- Gemini Omni Flash 1.1 clip cost on Sume, by length and resolution
What Gemini Omni Flash 1.1 costs on Sume per clip and per set of clips, from 360p to 4K, for each length it accepts. Native synced audio is included.
Written by Sume