Google brings Lyria 3.5 music generation to Gemini and the Gemini API

Google has moved its newest music generation model deeper into the Gemini ecosystem, announcing on September 4 that Lyria 3.5 is now available in the Gemini app and through the Gemini API. The release turns a research model from Google DeepMind into a more widely distributed creative tool, aimed not only at casual users who want to make a ringtone or birthday track, but also at creators, brands and developers building audio features into their own products.

According to Google, Lyria 3.5 is its best-sounding music generation model so far, with more expressive vocals and richer arrangements than earlier versions. In the Gemini app, users can describe a genre, choose between vocal and instrumental styles, use templates for common creative starting points and select shorter or longer track formats. The model is also being made available globally on Gemini’s web and mobile experiences, while artists and AI creatives can use it in Google Flow Music. Developers can access it through Google AI Studio and the Gemini API, and Google says the model is also available in Google Vids.

The announcement matters because AI music is moving from isolated demos toward everyday production workflows. For a video creator, a custom backing track can remove licensing friction. For a small business, a short jingle can be generated without hiring a studio for every variation. For developers, API access means the same model can become part of editing tools, campaign builders, games or education products where music needs to be generated around a user’s context.

Google is also entering a sensitive market. Music generation tools have drawn attention from artists, labels and regulators because they sit close to questions of copyright, consent and attribution. Google’s post focuses on availability and creative control rather than legal policy, so many details about rights management, training data and commercial use will still matter to professional users evaluating the model. The broader direction is clear, however: Gemini is becoming less of a text-only assistant and more of a multimodal production surface where writing, image, video and music tools sit inside the same user experience.

For the AI industry, Lyria 3.5 strengthens the argument that multimodal models will compete through workflow access as much as raw quality. A music model that lives in an app, an API, a creative suite and a video product can reach more people than one locked behind a standalone demo. Google’s challenge now is to make that reach useful without making the output feel generic or legally uncertain.