Summary: ElevenLabs released Music v2 on May 27, 2026, adding whole-song construction and mid-track genre changes such as opera to heavy metal and back. The company says the model was trained on licensed material and generated tracks are cleared for commercial use. Music v2 is available in ElevenMusic and ElevenCreative, while the API remains listed as coming soon.
ElevenLabs put out Music v2, the second generation of its music model, on May 27, 2026 — roughly ten months after the first version went live on August 5, 2025. The feature drawing every eye is genre switching: a single track can travel “from opera to heavy metal and back” and still hold together as one piece of music. The licensing story carries equal weight: the model was trained on licensed material and generated tracks are cleared for commercial use, drawing a clear line between ElevenLabs and Suno and Udio, both mired in litigation.
For content and marketing teams, AI music stops being a plaything and becomes stock material: scores for ads and branded video, produced without sync fees and without any clearance wait.
A Second Generation, Ten Months On
Short clips were all the first model could manage, which put a hard ceiling on its usefulness. Music v2 supplies the missing whole-song capability: you produce the piece section by section — intro, verse, chorus — and the model joins everything into one continuous work that keeps its structure and flow intact, rather than splicing together samples that have nothing to do with each other. ElevenLabs additionally claims broad gains across vocals, instrumentation, arrangement, and multilingual support, and says the model can carry rapid-fire rap delivery all the way through without falling apart.
Back in August 2025, when the first model shipped, ElevenLabs was already leading with commercial clearance rather than raw fidelity. Ten months on, that pitch has grown from “safe to use” to “safe, long-form, and steerable” — a fair measure of how quickly production demands have moved.
Switching Genres Inside a Single Track
The marquee feature is genre switching: partway through a track you call for a different style, and the model lands the change so it sounds like one continuous song rather than two glued together. Film and game scoring gain at once — the score follows a scene change on its own, with no editor hand-cutting between separate clips.
Producers get one more practical touch: non-musical sound effects can sit directly inside the generated track, so ambience or impact audio no longer demands its own overdub pass.
Build in Sections, Rewrite Just One
Inpainting lets you mark any section of a Music v2 track and regenerate that stretch alone from a prompt — swap out the bridge while the chorus stays exactly as it was. Music generation stops being a single dice roll and turns into an editing process you can repeat, feeling closer to cutting video or retouching photos than to pulling a slot-machine lever.
For teams shipping content at volume, iteration outweighs raw quality: a model you can steer section by section keeps compounding, and ten controllable generations beat a hundred left to chance.
Licensed Training Data, Commercial Use Cleared
ElevenLabs points out that Music v2 was developed “with artists, labels, and publishers,” extending rights-respecting licensing work that includes Believe. Set against 2026, the positioning carries real force: Universal and Sony want roughly 61,000 copyrighted works added to their Suno lawsuit after discovery revealed training on millions of their recordings, and Sony’s case against Udio has picked up more than 30,000 recordings.
Rivals spend their cycles in court; ElevenLabs sells certainty — a generated track you can place in a client campaign without wondering whether a takedown notice lands next quarter. For agencies and brands, the certainty itself is the product.
A Crowded Field, and Lower Prices
Competition in AI music intensified sharply across the first half of 2026. Google’s Lyria 3 Pro arrived March 25; Stability AI put out an audio model reaching six-minute songs on May 20; Suno has advanced to v5.5; and Flow Music, shown at Google I/O, took on covers, section editing, and music videos. The accelerating release rhythm is this year’s defining pattern — a thread our opening outlook for 2026 had already flagged.
ElevenLabs answered on capability and price at once: ElevenAPI now costs up to 50% less, and ElevenCreative self-serve plans are up to 40% cheaper. Music v2 can be used today in ElevenMusic — the consumer app that launched April 2, 2026 — and in ElevenCreative for marketing and branding work; the API remains tagged “coming soon,” with early access running through the sales team. The split is deliberate: ElevenMusic handles listening, remixing, and creating, while ElevenCreative aims at advertising and branded content — the exact segment where sync licensing has historically cost the most.
When developers and content teams assemble their AI music shortlist next quarter, the harder questions will be which licensing position survives legal review and which API is actually ready to build on — not simply which model sounds best.