Voice AI company ElevenLabs has unveiled Music v2, an upgraded music-generation model capable of changing genres within a single track. The new release arrives nearly 10 months after the company introduced its first AI music model and is designed to deliver more advanced compositions, stronger vocals, and greater creative control for users.
According to ElevenLabs, Music v2 can smoothly transition between very different musical styles, such as moving from opera to heavy metal and back again, while maintaining coherence. The model can also handle fast-paced rap vocals and incorporate non-musical sound effects directly into songs.
A major new feature allows artists to regenerate specific sections of a track using text prompts without affecting the rest of the composition. Instead of creating only short clips, users can now build songs piece by piece, including intros, verses, and choruses, before combining them into a complete production.
The company says the model performs more consistently across languages and delivers improved results for lyrics, vocals, and musical arrangements.
The launch comes as competition in AI-generated music continues to accelerate. Companies including Google, Stability AI, and Suno have recently introduced more advanced music models capable of producing longer and more sophisticated tracks. At Google I/O, Google also demonstrated tools for creating song covers, editing sections of music, and generating music videos through its Flow Music platform.
ElevenLab emphasised that Music v2 was trained on licensed data and is approved for commercial use, allowing users to use generated tracks in their projects freely. The company’s focus on licensed content comes as copyright concerns continue to surround the AI music industry, with competitors such as Suno and Udio facing legal scrutiny over training data practices.
Music v2 is available through ElevenLabs’ ElevenCreative platform and its recently launched ElevenMusic service, while support for ElevenAPI is expected to arrive soon.
