ElevenLabs Music generates background music from a written description, up to ten minutes. Call POST /v1/audio/speech with the description in `input`. Options: `duration_seconds` up to 600, `force_instrumental: true` to keep it wordless, `seed` to reproduce a take, and `music_model` to pick music_v1 or music_v2. Billing scales with the length produced. You can also steer it with a track you already have: upload one to POST /v1/music/references and pass the `reference_id` back here with an optional `reference_strength` (low–xhigh). It guides the sound, instrumentation, tempo and mood — it does not copy or remix the clip.
curl https://conduit-api.bifrostapi.net/v1/audio/speech \
-H "Authorization: Bearer ck-YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"eleven-music","input":"calm ambient game loop","duration_seconds":30,"force_instrumental":true}' \
--output music.mp3
# optional — steer it with a track you already have
curl https://conduit-api.bifrostapi.net/v1/music/references \
-H "Authorization: Bearer ck-YOUR_KEY" \
-F "model=eleven-music-reference" -F "file=@theme.mp3"
# → {"reference_id":"vLepwPUyEg5BQYZJxAF3_upload"} (10–600s clip)
curl https://conduit-api.bifrostapi.net/v1/audio/speech \
-H "Authorization: Bearer ck-YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"eleven-music","input":"same theme, battle version","duration_seconds":60,"reference_id":"vLepwPUyEg5BQYZJxAF3_upload","reference_strength":"high"}' \
--output battle.mp3→ 用 input 描述曲子,duration_seconds ≤ 600;加 force_instrumental:true 去人声。价格按时长计。参照音频时长须在 10–600 秒,上传一次收 2.10,之后拿这个 id 反复用不再收费;实际只取前 30 秒做参照。reference_strength 可填 low / medium / high / xhigh。
鉴权 · 计费 · 流式 → 快速接入