ElevenLabs AI music generative audio music generation API

ElevenLabs Music v2 Can Pivot From Opera to Metal Mid-Track and Somehow Stay Coherent

28 May 2026

ElevenLabs has launched Music v2, an upgraded AI music generation model capable of seamlessly transitioning between genres like opera and heavy metal, handling fast rap, and embedding sound effects within a single track. The model also features improved inpainting, allowing users to regenerate specific song sections independently, along with enhanced multilingual support. Trained exclusively on licensed data, all generated tracks are cleared for commercial use, with API pricing at $0.15 per minute.

ElevenLabs has quietly raised the bar for AI music generation with the release of Music v2, and the headline feature is exactly as chaotic as it sounds: a single generated track can move from operatic vocals to heavy metal without falling apart at the seams.

Beyond the genre-hopping party trick, the model brings genuine improvements across the board. Vocals are better, instrumentation is cleaner, and arrangements hold together more convincingly across styles. Fast rap is reportedly handled without the usual garbled mess, and sound effects can be embedded directly into tracks rather than bolted on afterwards.

The more practically useful upgrade is inpainting. You can now regenerate specific sections of a song without disturbing everything else. That might sound minor, but for anyone who has tried to fix a weak chorus while keeping a good verse intact, it matters. Building tracks section by section is now a first-class workflow rather than a workaround. Multilingual vocal support has also been improved, though ElevenLabs has been vague about exactly which languages got attention.

Music v2 sits underneath three separate products: ElevenMusic for musicians, the ElevenAPI for developers, and ElevenCreative aimed at brands and content teams. API access costs $0.15 per minute, with a five-minute maximum per generation and output at 44.1 kHz, 128-192 kbps. Serviceable quality, not exactly audiophile territory.

On the licensing front, ElevenLabs says the model was trained exclusively on licensed material, partly through a deal with music distributor Believe. Commercial use is cleared from the Starter+ plan upward, which should reduce the legal anxiety that tends to follow AI-generated audio around like a bad smell.

Whether the opera-to-metal transition holds up in practice or just looks good in a demo video remains to be seen.