ElevenLabs Music

ElevenLabs Music — Text-to-Audio Music Generation API

What is ElevenLabs Music?

ElevenLabs Music (Eleven Music) is a text-to-music model that turns a natural-language prompt, or a structured composition plan, into a complete studio-grade track — with real sung vocals and lyrics or fully instrumental. It understands both plain language and musical terminology, so a single prompt can set genre, tempo, key, instrumentation and song structure, and the model follows it. Version music_v2_5 is the newest tier, with better audio quality and prompt adherence than music_v2; music_v1 is the original prompt-only model. Tracks run up to 10 minutes and are built on licensed data cleared for commercial use.

Key Features

• •Finished songs with sung lead vocals and coherent lyrics, or instrumental-only via force_instrumental.
• •Multilingual vocals across English, Spanish, German, Japanese and more.
• •Section-by-section composition_plan (up to 30 chunks) for fixed structure and mid-track genre transitions.
• •Prompt-level control of genre, style, tempo and instrumentation.
• •Adjustable length (music_length_ms) and output encoding (MP3 or Opus up to 48kHz 192kbps).

Best Use Cases

Score a music video with a full vocal track, lay a v2.5 instrumental bed under a brand film, generate podcast and YouTube intros, or build ad variants and game background loops. Because it sings in multiple languages, it covers jingles and hooks that instrumental-only generators cannot. Reach for composition_plan when structure matters — labelled intro, verse and chorus sections, precise per-section length, or a genre change part-way through a track.

Prompt Tips and Output Quality

Name the genre, era and instrumentation, then describe how the song should move — for example a 1970s soul number that builds from a verse into a chorus with a key change. In testing, that prompt returned a full 60-second track at 48kHz with a clear build and a sung lead. Set force_instrumental to true for a clean backing bed. For transitions, put each genre in its own composition_plan chunk with its own positive_styles.

How to Use ElevenLabs Music

ElevenLabs Music turns a text prompt — or a section-by-section composition plan — into a complete, studio-grade track with vocals or instrumental. Write prompts the way you would brief a composer: name the genre, era, tempo, instrumentation and how the song should move.

Recommended Settings

• •model_id: music_v2_5 for the best quality and composition-plan support; music_v1 is prompt-only.
• •prompt: one or two sentences naming genre, mood, instruments and structure. Leave empty only when sending a composition_plan.
• •music_length_ms: around 60000 to fit a full verse, chorus and key change; the 30000 default is short for complete structure (prompt mode only).
• •force_instrumental: true for a vocal-free backing bed, false for songs with sung vocals.
• •output_format: auto (MP3 48kHz 192kbps on v2.5); pick an Opus option for smaller files.
• •composition_plan: up to 30 chunks, each with text (label in [brackets], directions in {braces}), duration_ms and positive_styles; seed reproduces a composition-plan generation.

Example prompts to try

Each example below ran end to end at the settings shown; the length and structure are exactly what the model returned.

• •Default — song with vocals (70s soul). Prompt: '1970s Philadelphia soul: a female lead belts the hook over lush string swells, a tight horn section, wah-wah guitar, Rhodes piano and a driving tambourine backbeat; the verse builds into a big chorus and a key change lifts the final chorus.' Returned a full 60-second track with a sung lead and a clear verse-to-chorus build; name the era and instrumentation and keep force_instrumental false for vocal-forward genres.
• •Multilingual vocals (Japanese). Prompt: '1980s Japanese city pop sung in Japanese: a smooth male lead over slap bass, clean funk guitar, bright FM electric piano and gated reverb drums; a breezy night-drive groove with a catchy chorus.' Sang a lead in Japanese over a funk groove; state the language explicitly to steer the vocal.
• •Mid-track genre transition (composition_plan). Leave the prompt empty and send a three-chunk plan — flamenco intro, a techno drop, then a flamenco return. Moved cleanly from solo guitar into four-on-the-floor techno and back; give each genre its own chunk and positive_styles.
• •Guaranteed instrumental (force_instrumental). Prompt: 'Propulsive instrumental heist-chase cue: fuzz bass, staccato brass stabs, live break-beat drums, clavinet and hand percussion; a tight funk groove that keeps accelerating.' Returned a bass-driven cue with no vocals; set force_instrumental true for a clean bed.

Try also 'reggaeton with a Spanish hook', 'a gospel choir build', or 'a slow bolero that switches to cumbia at the bridge'.

Can it make instrumental-only tracks?

Yes — set force_instrumental to true in prompt mode.

Which languages can it sing in?

English, Spanish, German, Japanese and more.

How long can a track be?

Up to 10 minutes.

What is a composition plan?

A JSON list of sections that replaces the prompt for precise structure and genre changes.

Which model version should I use?

music_v2_5 for the best quality and composition-plan support.