ElevenLabs sound effects API: 0.5-30 s, loop, and Sume's route

ElevenLabs' Create sound effect endpoint is POST https://api.elevenlabs.io/v1/sound-generation. It takes text (required), duration_seconds (0.5 to 30; if you leave it out, ElevenLabs guesses a length from the prompt), loop (a smooth loop, only on the eleven_text_to_sound_v2 model), and prompt_influence (0 to 1, default 0.3), and it returns an MP3. Sume does not list a sound-effect endpoint, so none of those four fields exist on Sume. What Sume does list for audio generation is the Music Router, which is a music surface with a different set of controls.

The ElevenLabs facts below come from its API reference page; the Sume facts come from the Music Router, Music 1.0 and Timeline 1.0 docs.

What does the ElevenLabs sound effect request take?

One required field and a few optional ones. The page lists output_format as a query parameter (codec, sample rate and bitrate, for example mp3_44100_128) and says that 192 kbps MP3 needs the Creator tier or above and 44.1 kHz PCM needs the Pro tier or above. The default model id is eleven_text_to_sound_v2, the only allowed value on that page.

Can I set a 2-second length on Sume?

Not with a field. Sume's docs say music length is steered in the prompt (for example a 2-minute track or section markers such as [0:00-0:30] Intro: ...), and that duration and duration_seconds are unrecognized and rejected. The docs describe full songs up to a few minutes, so the Music Router is built for tracks, not for sub-second hits. If you need an exact short cut, generate a track and trim it with a Timeline render.

How do I loop a short sound under a video on Sume?

Use the Timeline soundtrack object. Its fields are url, gain_db, loop, fade_out_seconds (up to 10) and duck_db (0 to 20). The OpenAPI description says loop repeats the bed until the spine ends; without it a soundtrack shorter than the spine stops partway through and the render warns soundtrack_shorter_than_spine. That loops a file you already have, including one generated by a music job. It does not synthesize a seamless loop the way ElevenLabs' loop flag does, so listen for the seam.

What should I do today?

If your product needs discrete effects (a click, a whoosh, a door), use a sound-effect service for those and Sume for the music bed and the render. If the video model you use already generates sound, that is a separate route covered in AI sound effects for video. Check the generated audio before you ship it; Sume's docs call prompt directions creative guidance, not guaranteed settings.