Google has introduced a new music-generation capability to the Gemini app, expanding its generative AI tools beyond text and images. The feature, currently in beta, is powered by DeepMind’s latest Lyria 3, designed to produce short, original music tracks with lyrics based on user prompts.
Users can describe the desired genre, mood, tempo, or vocal style, while also uploading photos or videos to guide the emotional tone of the composition. The system generates up to 30-second tracks, complete with AI-created cover art, offering more nuanced and realistic sound than previous versions. According to Google, Lyria 3 allows finer control over musical elements and supports more complex arrangements.
Alongside the Gemini rollout, the model is becoming available to creators on YouTube through the Dream Track feature, which is now expanding globally after a limited U.S. release. While direct imitation of specific artists is restricted, prompts referencing artists are interpreted as general stylistic inspiration rather than exact replication.
All music generated with Lyria 3 includes a SynthID watermark, enabling identification of AI-created audio. Gemini is also gaining tools to detect whether uploaded tracks were generated by AI. The feature is rolling out worldwide to users aged 18 and over, with support for multiple major languages.