AI sound effects generator API: an open-weight SFX model, then Sume

Sume's docs list no sound-effects generator API. A model built for that job is open-weight: Stability AI says Stable Audio 3.0 Small SFX is on Hugging Face. If you want effect files today, run that model yourself, then bring the files into Sume and layer them in Timeline 1.0. Sume's own audio routes are music and speech.

The Stable Audio fact is from Stability's announcement, read 2026-09-29. The Sume facts are from the Music Router and Timeline 1.0 docs. We did not run Small SFX, so this post does not describe its output.

What does Sume list for audio?

How do I get my own effect files into a Sume render?

Timeline 1.0 accepts only Sume-hosted media.sume.com files. Its docs say to import first with POST /v1/media-imports, and to send an Idempotency-Key on the render. Once imported, an effect file is a normal Sume audio file you can use as the render's soundtrack bed, which takes url, gain_db and loop.

Can I ask the music route for an effect?

You can write any description into a Music Router prompt, but the route makes a music track, priced at the fixed Music price per generation. It is not a substitute for a one-shot effect, and Sume's docs make no promise about isolated sounds. Listen before you rely on it.

What should I check before using an open-weight SFX model?

• Read the model's license on its own page. Stability names a Community License and an Enterprise License for organizations with $1M or more in revenue.
• Check the file format and length your render needs before importing.
• Keep the source model and prompt in your project notes, as you would for any generated asset.

What can the Timeline soundtrack bed do?

The soundtrack field takes a Sume-hosted url and optional gain_db, loop, fade_out_seconds (up to 10) and duck_db (0 to 20). Ducking lowers the bed under the main audio and needs a real audio spine, not silence. A short effect loops or fades like any other bed, so a single imported file can be placed once or repeated.

Timeline 1.0 charges $0.10 per output minute, rounded up, and its docs say no provider inference is involved, only worker ffmpeg.

What is the smallest workflow that works?

Generate or pick an effect file outside Sume, import it with POST /v1/media-imports, and reference its media.sume.com URL in a render's soundtrack. Run POST /v1/timeline-1.0/plan first: it is an unbilled compile preflight that checks the request before you pay for a render.