Google has made its Lyria 3.5 music-generation model available through the Gemini app, Google AI Studio and the Gemini API. The release is available globally to Gemini users on the web and mobile, placing AI music creation inside one of Google’s largest consumer AI products.
For producers, the important change is distribution rather than a new standalone music app. Lyria 3.5 can now be accessed through a prompt-based Gemini workflow, while AI Studio and API access create routes for developers to test the model and build it into other creative tools.
Key Takeaways
- Lyria 3.5 is now available in the Gemini app globally on web and mobile.
- Google has also opened access through Google AI Studio and the Gemini API.
- Gemini users can describe or select genres and choose vocal or instrumental styles, with templates intended to provide starting points.
- Google says the model produces more expressive vocals, richer arrangements and higher-fidelity tracks than previous versions; those are the company’s claims rather than independent benchmarks.
- Google’s release material does not set out Lyria 3.5’s training data or the licensing terms that may apply to generated music.
What Lyria 3.5 adds to Gemini
According to Google’s Lyria 3.5 announcement, users can choose or describe a genre, select vocal or instrumental output, use templates and request short or longer tracks. Google positions the feature for uses such as backing tracks, jingles, personalised songs and ringtones.
That makes Lyria less of a specialist experiment and more of an accessible sketching tool. A producer could use text prompts to generate a rough arrangement reference, explore a genre direction for a brief, or make temporary music for an edit. It should not be confused with a DAW workflow: nothing in the announcement suggests that Gemini generates multitracks, exports stems or offers the detailed editing expected when arranging, mixing or finishing a release.
The model is described by Google as its best-sounding music generator so far, with improvements to vocal expression and arrangement richness. The company’s public Gemini music-generation page is the practical destination for users wanting to check how the feature is presented in the app.
API access is the producer-tool story
The Gemini app rollout gives Lyria 3.5 a broad user base, but API availability may prove more relevant to music-software developers. Access through the Gemini API means third parties can potentially incorporate AI music generation into their own services, rather than asking creators to leave a production environment and work directly in Gemini.
That could support prompt-to-demo features in songwriting tools, video-editing platforms, content-creation services or educational music applications. For production-software companies, the model is another option for building assisted composition or music-for-picture functions alongside existing generative-audio providers.
Google has also made the model available through AI Studio, offering a more developer-oriented route for experimentation. Developers can refer to Google’s Lyria 3.5 music-generation documentation for the current implementation details and API requirements.
For working producers, the distinction is useful. Gemini is the immediate route for trying the feature; API access is not a replacement for a studio tool unless another developer builds a workflow around it. Whether future integrations offer DAW-friendly controls, export options or more precise arrangement editing will determine its practical value in serious sessions.
Rights questions remain outside this announcement
Google’s rollout arrives while AI music generation remains closely tied to copyright and training-data debates. The announcement promotes creative controls and output quality, but does not explain the music used to train Lyria 3.5 or establish the ownership, commercial-use and distribution conditions for generated tracks.
That leaves several questions producers should resolve before putting any AI-generated material into a commercial release, label pitch, sample pack or client deliverable. Check the applicable product terms, establish whether a platform permits monetised use, and keep a record of prompts and generated versions where a project has clearance requirements.
It is also sensible to treat generated vocals, lyrics and stylistic prompts carefully. A tool may be useful for ideation or a temporary reference while still being unsuitable for a final master where originality, credits and rights-chain documentation matter.
What to watch next
The immediate story is that Google has moved Lyria 3.5 from a narrower AI-music context into Gemini’s mainstream interface while keeping a developer route open. Producers can test the consumer experience now, but the more consequential development may be the tools built around the API.
The next useful details will be Google’s terms for generated music, any clearer information on training and attribution, and whether Lyria 3.5 reaches production environments with controls that go beyond generating a finished stereo track.