ElevenLabs Music v2 generates complete studio-quality songs — vocals, instrumentation, and arrangement — trained only on licensed data and cleared for commercial use. Regenerate any verse, chorus, or bridge without touching the rest.
Built exclusively on licensed music through deals with Merlin and Kobalt, so every track is cleared for commercial use — no sync fees, no clearance delays.
Select any part of a track and regenerate just that section. Rework the bridge without touching the chorus, or rebuild an intro from a prompt.
Shift genres inside a single song — opera to metal and back in a handful of bars — while keeping vocals and structure coherent.
Generate sung vocals with custom lyrics in English, Spanish, German, Japanese and a growing list of languages, or produce instrumental-only tracks.
ElevenLabs Music v2 is the second generation of ElevenLabs' AI music model, released in late May 2026 under the Eleven Music product. From a text prompt it generates a complete song — vocals with custom lyrics, instrumentation, and full arrangement — or an instrumental-only track. Its defining trait among major AI music tools is that it is trained exclusively on licensed catalog data, secured through partnerships with the independent-label alliance Merlin and publisher Kobalt, which makes generated tracks cleared for commercial use on paid plans.
Beyond raw generation, v2 focuses on control and editability. Improved inpainting lets creators regenerate any single section of a track — a verse, chorus, or bridge — without disturbing the rest, and the model can switch genres mid-song while keeping vocals and structure coherent. It supports multilingual vocals, embedded non-musical sound effects, and stem separation for downstream mixing. The model is available in the ElevenLabs app and, for paying subscribers, through the Eleven Music API.
From a single text prompt, Music v2 produces a complete track with sung vocals, instrumentation, and arrangement. You can supply your own lyrics or let the model write them, and generate instrumental-only versions when no vocals are needed.
The upgraded inpainting engine lets you highlight any region of a track and regenerate only that part. Rework a weak bridge, swap a chorus melody, or rebuild the intro while the rest of the song stays exactly as it was.
A signature capability of v2: the model can transition between genres within a single song — for example moving from opera to metal and back over a few bars — while maintaining vocal coherence and structural continuity.
Music v2 sings in multiple languages including English, Spanish, German, and Japanese, with a growing roster of supported languages. Lyrics and pronunciation are handled per language rather than transliterated.
Because the model is trained only on licensed data via Merlin and Kobalt agreements, tracks generated on paid plans are cleared for commercial use without separate sync licensing. Free-tier output carries usage restrictions.
Downloads can be delivered as a full mix, split into vocals and instrumentals, or separated into individual stems, giving producers material to remix, master, or drop into a DAW.
Creators can steer tempo and musical key and compose a song section by section — building verse, chorus, and bridge independently — for tighter control than a single one-shot generation allows.
| Model | ElevenLabs Music v2 |
|---|---|
| Input | Text prompt (+ optional custom lyrics) |
| Output | Vocal songs or instrumental tracks |
| Max length | Up to 10 minutes (API); ~5 minutes (app) |
| Minimum length | ~3 seconds |
| Inpainting | Regenerate individual sections |
|---|---|
| Genre switching | Mid-track transitions supported |
| Structure | Section-by-section composition (verse/chorus/bridge) |
| Controls | Tempo, key, prompt-guided regeneration |
| Sound effects | Non-musical SFX can be embedded |
| Vocals | Multilingual (English, Spanish, German, Japanese, +more) |
|---|---|
| Stems | Full mix, vocals/instrumental split, or 4 stems |
| Training data | Licensed only (Merlin, Kobalt) |
| Commercial use | Cleared on paid plans; restricted on free tier |
| Interfaces | ElevenLabs app; Eleven Music API |
|---|---|
| API access | Paid subscribers |
| Released | May 2026 |
Because paid-plan output is cleared for commercial use, marketers and video editors can score ads, product videos, and brand content without chasing sync licenses or worrying about takedowns.
YouTubers, podcasters, and streamers can generate custom, on-brand tracks in the length and mood they need, then regenerate sections that don't quite fit rather than starting over.
Musicians can sketch full songs with vocals and lyrics, iterate on a specific verse or chorus via inpainting, and export stems to finish the production in their own DAW.
Developers can produce genre-flexible tracks — including mid-track transitions for shifting scenes — and pull stems for adaptive, layered audio systems.
Teams releasing content across markets can generate vocal tracks in several languages from the same creative direction, keeping tone consistent while localizing the lyrics.
Music v2 is a broad quality and control upgrade over the first Eleven Music model, adding mid-track genre switching and much stronger section editing.
| Feature | ElevenLabs Music v1 | ElevenLabs Music v2NEW |
|---|---|---|
| Released | Aug 2025 | May 2026 |
| Vocal & arrangement quality | Strong | Broad quality jump |
| Section inpainting | Basic | Improved, prompt-guided per section |
| Mid-track genre switching | Not featured | Yes |
| Multilingual vocals | Supported | Improved, more languages |
| Licensing | Licensed data, commercial-cleared | Licensed data, commercial-cleared |
Full commercial clearance applies to paid tiers (from the Starter plan up). On the free plan, commercial use is heavily restricted or removed, so read the plan terms before deploying tracks publicly.
The Eleven Music API is available to paying subscribers, and v2 API access rolled out after the app. Availability of specific v2 endpoints and features can vary by plan and region.
Maximum track length differs between the app (~5 minutes) and the API (up to 10 minutes), and very long, highly structured compositions may need section-by-section assembly rather than a single generation.
As with all generative music, output quality varies by prompt and genre, and complex vocal or lyrical requests may need several iterations. A higher-tier Eleven Music Pro model was announced as a later release.
We're adding ElevenLabs Music v2 as platform access opens — explore SharkFoto's available AI tools today and check back for availability.
Try ElevenLabs Music v2 now