ElevenLabs Music V2
Full music tracks from a text description — any genre, with vocals or fully instrumental. Pick the exact length, from a 3-second sting to a 10-minute piece. Output is 48 kHz / 192 kbps MP3, royalty-free.
Capabilities
| Feature | Support |
|---|---|
| Text-to-Music | Yes |
| Duration | 3 seconds – 10 minutes (you set the exact length) |
| Vocals | Yes (or explicitly instrumental) |
| Output | 48 kHz / 192 kbps MP3 |
| Licensing | Royalty-free output |
| Reference audio / image | No |
Exact-Length Generation
Unlike Lyria (fixed 30s clip or ~2min song), Music V2 renders to the exact duration you request. That makes it the go-to for:
- Scoring to picture — a 47-second cue for a 47-second cut, no trimming.
- Long-form beds — podcast intros, ambient loops, 10-minute focus tracks.
Prompting Tips
- Genre + mood + instrumentation + tempo is the reliable skeleton: "cinematic orchestral, tense build, low strings and taiko, 100 BPM".
- Structure long tracks explicitly. For multi-minute pieces, outline sections: "opens sparse, drums enter at one third, full arrangement by the midpoint, stripped-back outro".
- Say "instrumental only" when you don't want vocals — the model may otherwise add them where the genre implies it.
- Lyric themes work — "an upbeat indie-pop song about leaving home" steers the generated lyrics.
Limitations
- No reference-audio conditioning (can't continue or match an existing track)
- No stems — output is a single mixed file
- No seed / determinism
See Also
- Lyria 3 — Google's music family with timestamped structure control and image references
- ElevenLabs Sound Effects — same provider, SFX instead of music