The AI music race just got a lot louder. On May 28, 2026, voice and audio company ElevenLabs released Music v2, a major upgrade to its AI music generator that can travel from opera to heavy metal and back inside a single track without the song ever falling apart. It is one of the most ambitious creative AI releases of the spring, and it arrives at a moment when nearly every major lab is fighting to own generative audio. For musicians, content creators, and marketers, it could be the most practical AI music tool released so far this year.
What is ElevenLabs Music v2?
Music v2 is the second generation of the AI music model ElevenLabs first launched roughly ten months ago, back in August 2025. Like its predecessor, it turns a simple text prompt into a full, studio-quality song complete with vocals, instrumentation, and arrangement. The difference this time is polish and control. ElevenLabs says the new model delivers noticeably better vocals, richer instrumentation, and more convincing arrangements across every genre it touches, and that it performs far more reliably across different languages, lyrics, and vocal styles. In short, it is less of a novelty and more of a tool you could actually finish a project with.
Switching genres mid-track is the headline feature
The standout trick, and the one earning Music v2 most of its attention, is its ability to change genres in the middle of a song while keeping everything musically coherent. ElevenLabs demonstrated a single track moving from soaring opera into heavy metal and back, handling fast, dense rap without slurring the words, and even weaving in non-musical sound effects. That kind of stylistic range has been the weak spot for AI music up to now, where most models lock into one mood and stay there. Being able to engineer a dramatic genre flip on purpose opens the door to trailer music, dynamic game scores, and the kind of attention-grabbing transitions that perform well on social video.
Section-by-section editing and smarter inpainting
The bigger workflow story is control. Earlier AI music tools forced you to regenerate an entire song just to fix one weak bar, which made fine-tuning painful. Music v2 introduces improved inpainting, so you can highlight a single section and regenerate just that part while the rest of the track stays exactly as it was. You can also build a song the way a real producer would, one piece at a time.
- Section-based composition: construct a song from individual parts such as the intro, verse, and chorus, then stitch them together into a finished piece.
- Targeted inpainting: rewrite or restyle one section using a prompt without disturbing the surrounding music.
- Stronger multilingual support: more dependable results across languages, lyrics, and vocal performances.
- Embedded sound effects: drop non-musical audio cues directly into a generated track.
Built on licensed data and cleared for commercial use
For working creators, the most important detail may not be the music at all, but where the training data came from. ElevenLabs says Music v2 was trained only on licensed material, in part through a partnership with the music company Believe, and that every track it generates is cleared for commercial use starting at the Starter+ plan. That matters because rival AI music startups such as Suno and Udio have been tangled in copyright lawsuits from major labels over how their models were trained. A model you can confidently use in a monetized YouTube video, a client ad, or a paid podcast without worrying about a takedown is a genuinely different proposition, and it is quickly becoming the dividing line in this category.
Pricing and where you can use it
Music v2 powers three different ElevenLabs products. ElevenMusic is the consumer-facing app for musicians and hobbyists, ElevenCreative targets marketing and branding teams, and the ElevenAPI lets developers build the model into their own apps. API pricing lands at about $0.15 per minute of generated audio, with tracks capped at five minutes and output at 44.1 kHz and roughly 128 to 192 kbps. That is a reasonable rate for short-form content, social clips, and background scoring, even if album-length projects will need some stitching. There is a free tier to experiment with, but commercial rights kick in once you move up to a paid plan.
How Music v2 fits into the 2026 AI music race
Music v2 did not land in a vacuum. The last few months have turned into an all-out sprint for professional-grade AI audio. Google launched its Lyria 3 Pro model and, at Google I/O 2026, gave its Flow Music tool the ability to create covers, edit songs by section, and generate full music videos. Stability AI rolled out an audio model capable of six-minute songs, and Suno pushed out its v5.5 update. Against that backdrop, ElevenLabs is betting that flawless genre control plus airtight commercial licensing will set it apart from competitors chasing raw length or spectacle. We have been tracking each of these launches closely over on the Creative AI Show, and the pace of releases in generative music right now is genuinely hard to keep up with.
The takeaway for creators
ElevenLabs Music v2 is less about flashy demos and more about removing the friction that has kept AI music out of serious projects. Section-level editing makes it controllable, the licensed training data makes it safe to monetize, and the genre-switching gives creators a creative hook they could not get anywhere else. If you make videos, run a podcast, or produce content for clients, this is one worth testing before your competitors do. For more hands-on breakdowns and demos of the newest creative AI tools, subscribe to our YouTube channel and follow along on the Creative AI Show.
No responses yet