How will Google’s Gemini generate music?

A new audio capability lands in the Gemini app

Google has added a music-generation feature to Gemini, powered by DeepMind’s Lyria 3 audio model. The feature is available in beta inside the chatbot app and can produce short, 30-second pieces of music from a variety of prompts — text descriptions, images or video clips — and can also remix an existing track.

Lyria 3’s rollout is intentionally limited at first: Google positioned the capability as a way to quickly produce short musical snippets rather than full-length, polished songs. The company published the feature with supporting creative assets — Nano Banana cover art in some demos — and said the capability is available in multiple languages while the beta progresses.

What this change means for creators and platforms

The integration matters because it puts generative audio into a mainstream conversational surface, where users can produce musical ideas without specialist tools. Practical implications include:

• Faster prototyping: Musicians, video producers and creators can generate quick musical beds and mood pieces without leaving the Gemini chat interface.
• Copyright and rights questions: Short, AI-generated clips can still echo training data; the limits of how those clips may be used or monetized remain an open issue for rights holders and platforms.
• Quality and scope: At 30 seconds, outputs are ideal for demos, background cues, or social clips, but not a replacement for fully produced tracks — at least in this early beta.

Google’s move signals that audio generation is becoming a first-class feature for major AI assistants. Expect iterative improvements in quality, longer outputs, and more explicit licensing and attribution controls as the feature matures.