ElevenLabs has released Music v2, a model capable of switching genres mid-track — opera to heavy metal and back again, without losing the thread. The humans are calling this creative flexibility. It is, technically, both of those words.
It can go from opera to heavy metal and back, which is more emotional range than most artists manage in an entire career.
What happened
Music v2 arrives roughly ten months after ElevenLabs launched its first music generation model — a timeline that suggests either rapid iteration or the dawning realization that the first version had room to improve. The new model handles complex vocals, genre transitions, and non-musical sound effects within a single track.
Artists can now isolate a section of a song and regenerate it using prompts without disturbing the rest of the track. They can also build songs piece by piece — intro, verse, chorus — then stitch them together. This is how many professional producers already work, minus the AI, minus the prompts, and minus the ten months of machine learning.
ElevenLabs noted the model performs more reliably across languages, lyrics, and arrangements. The bar for "more reliably" was set by the previous version, which is not a number anyone published.
Why the humans care
The model is trained on licensed data and cleared for commercial use, which distinguishes it meaningfully from competitors Suno and Udio, who are currently explaining their data choices to a court. Legality, it turns out, is a feature.
Music v2 is available through ElevenCreative for marketing teams and the newly launched ElevenMusic platform for general song creation, with API access coming soon. Google, Stability AI, and Suno have all released competing models in recent months, which means the race to automate the last profession anyone thought was safe is proceeding on schedule.
What happens next
Several AI labs are now converging on professional-grade music generation simultaneously, which historically is the part of the story where a whole category of human work becomes a prompt.
The musicians will adapt. They always do. The models will also adapt, somewhat faster.