Google DeepMind's Gemini app rolls out Lyria 3 music generation

The new feature, accessible via the “Tools” menu within the Gemini app, leverages the Lyria 3 model to translate user prompts into 30-second audio tracks. According to Google DeepMind, the model provides users with greater creative control, allowing them to specify elements like musical style, vocal type, and tempo. The output is accompanied by custom cover art generated by another AI system, Nano Banana.

This advancement in autonomous lyric generation marks a significant technical and philosophical shift from previous AI music tools, which typically required users to supply their own words. The development lowers the barrier to entry, effectively enabling anyone with an idea to function as a complete songwriter. Google DeepMind states the model is “designed for original expression, not for mimicking existing artists,” a declaration that places it in direct opposition to other services that have faced criticism for generating vocal clones of popular musicians.

In a clear effort to preempt copyright concerns that have plagued other generative AI sectors, Google has implemented a multi-layered approach. The company states it was “very mindful of copyright and partner agreements” during the model’s training. If a user specifies a particular artist in their prompt, the system will treat it as “broad creative inspiration” to create a track in a similar style, rather than attempting to replicate the artist’s voice or specific works. Furthermore, Google has installed filters to check outputs against existing content and has provided a mechanism for users to report potential copyright violations, acknowledging that its approach “might not be foolproof.”

The functionality extends beyond text prompts. Users can also upload a photo or video clip, and the AI will compose a track with lyrics tailored to the visual content’s mood. This multimodal approach, which builds upon Gemini’s existing capabilities with images and video, suggests a strategy of creating a comprehensive, all-in-one generative media suite within a single application.

The rollout of Lyria 3 represents Google’s most assertive entry into the competitive AI music generation space, challenging incumbents and startups alike. By embedding this capability directly into its flagship consumer AI app, Google is betting on accessibility and integration to drive adoption. The move also signals a broader industry trend where major technology platforms are rapidly consolidating advanced generative AI features into their core products, transforming them from simple interfaces into powerful creation engines.

What remains to be seen is how the market will respond to a tool that explicitly avoids the viral, if ethically fraught, ability to mimic famous artists. Google’s cautious, rights-first stance could be a differentiating factor that attracts certain users and enterprise partners, or it could limit the tool’s appeal in a landscape where user demand often leans toward imitation. The feature is currently in beta, and its evolution will be closely watched as both a technical marvel and a case study in the complex interplay between AI innovation and intellectual property rights.