Google is rolling out its Lyria 3 AI music model to the Gemini app today, marking a major expansion of AI music generation capabilities. The new model, developed by Google DeepMind, offers faster generation times and improved quality compared to previous versions. Users can now create 30-second musical snippets by simply describing what they want or uploading an image to set the mood.

Unlike earlier versions, Lyria 3 doesn't require users to provide lyrics—the model generates appropriate lyrics automatically based on the prompt. Each creation comes with an album cover-style image generated by the Nano Banana model. The feature is accessible through a new "Create music" option in both the Gemini mobile app and web interface, with pre-loaded tracks available for remixing.

Google has also announced that Lyria 3 will integrate with its Dream Track toolkit for YouTube Shorts, complementing the platform's existing Veo AI video generation capabilities. The company provided sample prompts demonstrating the model's versatility, from Afrobeat tracks about plantains to 1970s Motown soul and sea shanties. This move positions Google as a major player in the growing AI music generation space, where several companies already offer similar tools that have flooded streaming platforms with AI-generated content.