ElevenLabs launches Music v2, its new music generation model

ElevenLabs launches Music v2, an upgraded music generation model with refined multilingual vocals, section-by-section composition, and API price cuts.

ElevenLabs launches Music v2, the new version of its music generation model. The update aims for more refined vocals, instrumentation, and arrangements across all genres, with expanded multilingual support for both lyrics and vocals.

In terms of control, the model handles a new level of vocal and compositional complexity: a single track can switch from one genre to another, from opera to metal, incorporate fast rap, dense lyrics, or non-musical sound effects without losing coherence. Inpainting allows users to select a specific portion to regenerate it alone, for example, reworking the bridge while leaving the chorus intact. Composition also occurs section by section—intro, verse, chorus—allowing for the creation of a full track rather than short snippets.

Music v2 powers three services with distinct uses: ElevenMusic for listening, remixing, and creating; ElevenCreative for downloadable music intended for advertising and brand content; and ElevenAPI for integration into a product, with the latter expected soon. The company is also taking the opportunity to reduce its prices, up to 50% on the API and up to 40% for self-service on ElevenCreative.

Finally, ElevenLabs emphasizes the rights aspect: the model is reportedly trained solely on licensed data and validated for commercial use, without synchronization fees or clearance delays, as an extension of its collaboration with Believe.