Google has given its virtual assistant Gemini the ability to create AI-generated music.
The tech giant is testing its “most advanced music generation model yet”, Lyria 3, which is now available via Gemini.
According to the site, Lyria 3 allows fans to “express, explore, and experiment with high-fidelity music, using prompts to create tracks with natural flow from note to note”.
“We’ve developed it with input from producers and musicians so it understands musicality – from rhythm to arrangement,” Google added.
Users are instructed to “describe” their track, “then dial in the details” they want. “Lyria 3 handles the complexity, putting new musical possibilities at your fingertips,” it continued.
Lyria 3 lets fans create “cohesive” songs, “explore global languages and genres”, and compose music by images by uploading photos. The model promises that the finished result will be “professional-grade audio”.
All tracks generated in the Gemini app are “embedded with SynthID, our imperceptible watermark for identifying Google AI-generated content”.
There are currently concerns online over what Lyria 3 was trained on to create music. Speaking about the nature of the model’s training, a representative told Billboard that it is “mindful of copyright and partner agreements”. They said Lyria 3 only trains on music that YouTube and Google have “a right to use under our terms of service, partner agreements, and applicable law”.
Google just added Lyria, its music generator, to Gemini.
What was it trained on? They haven’t said.
They say they have been ‘mindful of copyright’. What does this mean?! It sounds worryingly like intentional misdirection.
You should assume it’s trained on copyrighted work… pic.twitter.com/0ceawY9K7u
— Ed Newton-Rex (@ednewtonrex) February 18, 2026
On X/Twitter, one person said this “sounds worryingly like intentional misdirection”, adding: “You should assume it’s trained on copyrighted work without a licence unless they come out and say it’s not.”
You can find more information on the new model here.
Google launched its Project Genie creation tool last month, which has seen users rip off hit games like The Legend Of Zelda, Mario and Grand Theft Auto.
Earlier this month, Oliver Schusser, Apple Music’s Vice President, said the platform had demonetised two billion “fraudulent” music streams worth around $17million in royalties last year (via MixMag).
Deezer recently announced that it has demonetised 85 per cent of all AI-generated tracks on its site using an AI-detection tool – which it is now making commercially available for others to purchase. Last September, Deezer said 28 per cent of music uploaded to the platform was fully AI-generated.
Bandcamp has banned all AI-created tracks, too, saying: “We reserve the right to remove any music on suspicion of being AI-generated.”
Other critics of the tech include Mac DeMarco and SZA, while ABBA‘s Björn Ulvaeus has called AI “such a great tool”. Elsewhere, Kehlani hit out at an AI-generated artist receiving a $3million (£2.2million) record deal.
Last November, Teddy Swims admitted that he sometimes uses the tech to make music: “If you use it the correct way, I think it’s a beautiful tool.” At the end of 2025, Jorja Smith’s record label criticised a song made using AI, alleging that it had “cloned” her voice.
