elevenlabs-music-generation

Run the runcomfy CLI with the elevenlabs/elevenlabs/music-generation model, passing a JSON input with a prompt field containing your style description and lyrics. The CLI submits the request, polls for completion, and downloads the audio to your output directory.

Set force_instrumental to true in the input JSON and also state "no vocals" in the prompt text. This produces instrumental-only output suitable for podcast intros, study playlists, and game loops.

ElevenLabs Music supports 5 seconds to 5 minutes per call, controlled by the music_length_ms parameter (5000-300000). For longer pieces, generate sections separately and stitch them externally.

Pricing is approximately $0.0083 per second of generated audio, so 30 seconds costs about $0.25 and 5 minutes about $2.49. Draft at short durations to validate direction before paying for full-length renders.

No, this endpoint only offers a force_instrumental toggle and cannot request specific voice identities or clone singers. Voice cloning is a separate ElevenLabs capability not exposed through this music-generation endpoint.

Awkward phrasing usually comes from inconsistent lyrical meter in the prompt. Keep even syllable counts per line and a clear rhyme scheme, since the model follows the meter you provide.