Google has added music generation to its Gemini app, allowing users to create 30-second tracks from text prompts or even images. Joël Yawili and Myriam Hamed Torres write in the Google Blog that the feature is powered by Lyria 3, Google DeepMind’s latest generative music model, and is currently available in beta.
Users can describe a genre, mood, or memory, and the model generates a track complete with lyrics and cover art. Alternatively, users can upload a photo or video, and Gemini composes a fitting soundtrack based on the visual content. The model also handles lyric writing automatically, without requiring user input.
Compared to its predecessors, Lyria 3 offers more creative control over style, vocals, and tempo, and produces more musically complex results. Tracks are generated in seconds and can be shared via download or a link.
All generated audio is embedded with SynthID, Google’s watermarking system for AI-generated content. Users can also upload audio files to the Gemini app to check whether they were created using Google AI.
Google states that the model is designed for original expression and will not directly replicate a named artist’s style. Filters check outputs against existing content, and users can report potential rights violations.
