ElevenLabs ships SFX v3: 30s multi-event sound design via API
The limited-access model sequences a whole scene in one prompt, loops seamlessly, and now does music and vocals—30s clips, 40 credits a second.
Copy markdown
Sequence a whole scene in one call
SFX v3 (model `eleven_text_to_sound_v3`) follows multi-part prompts like “footsteps on gravel, then a metallic door opens,” so you generate layered sound design in a single API request instead of stitching clips. It’s live on the text-to-sound-effects endpoint; v2 stays the default.
30-second clips that loop seamlessly
Set any duration from 0.1 to 30 seconds at 40 credits per second, and flag longer beds to loop for gapless playback—enough to score game ambience, UI sounds, or app background loops straight from code, no editor.
It can sing now, and the highs are clean
v3 handles music and vocal prompts far better than v2 and cuts the clipping on chimes, gongs, and whistles, making it usable for stingers and jingles, not just foley. Catch: it’s limited-access and region-gated—accounts without it get an “Unknown model” error.