ElevenLabs released a music generation model that allows users to change genres partway through a track, the company announced Tuesday.
The model also lets users regenerate a specific section of a song while leaving the rest of the audio intact — a capability that separates it from most existing AI music tools, which typically require users to regenerate an entire composition to change any part of it.
What the model does
The section-level editing feature works like a non-destructive audio tool, allowing targeted changes without disturbing surrounding material.
Genre switching mid-track opens the door to compositional structures that blend styles — a verse in one genre resolving into a chorus in another, for example — without the user needing to stitch together separately generated clips.
ElevenLabs has not yet published a peer-reviewed technical paper or third-party benchmark data describing the model’s architecture or training set.
ElevenLabs’ position in AI audio
ElevenLabs built its early reputation on AI voice cloning and text-to-speech synthesis before expanding into music and sound effects.
The company reached a $1.1 billion valuation following a $80 million Series B funding round in January 2024, according to Reuters. It has since moved aggressively into generative audio beyond speech.
The AI music generation sector has grown crowded. Competitors include Suno and Udio, both of which offer text-to-music generation, as well as Google’s MusicLM and Meta’s AudioCraft — tools developed by two of the largest technology companies in the world.
Still, none of those tools have publicly demonstrated mid-track genre switching with section-level regeneration as a core user-facing feature.
Industry context
The broader generative AI music market faces unresolved legal questions. In June 2024, a coalition of major record labels — including Universal Music Group, Sony Music, and Warner Music Group — filed copyright infringement lawsuits against Suno and Udio in U.S. federal court, according to AP.
Those cases remain active and could shape how AI music companies license or restrict training data going forward.
ElevenLabs has not publicly disclosed the dataset used to train its new music model, nor whether it secured licenses from rights holders for that material.
The Recording Industry Association of America has pushed for stricter disclosure standards from AI developers using copyrighted audio for model training.
Music producers and independent artists have begun using AI generation tools to produce demo tracks, score short-form video content, and prototype arrangements before committing to full productions.
The global music production software market was valued at $8.5 billion in 2023, according to a report by Statista, with AI-assisted tools representing a growing share of that figure.
ElevenLabs has not announced a standalone release date for the music model or confirmed whether it will be integrated into the company’s existing product platform under a separate pricing tier.
