Google DeepMind has released an updated version of its music generator Lyria 3, and now it is built right into the Gemini chat-bot interface. Neureset is able to convert not only text but also uploaded pictures or videos into audio - just give a description, and the system will create the track itself.

In the new version, the developers highlighted three key points. Lyria 3 now writes lyrics independently based on request, gives the user enhanced control over style, vocals and speed, and can deliver compositions with high musical complexity. Each roller’s time frame is 30 seconds. The company emphasizes: the task is not to create hits, but to give people another tool for creativity.

Now available in the desktop version of Gemini, and soon to be available on mobile. Users with Google AI Plus, Pro, and Ultra subscriptions will receive extended limits per generation. All generated tracks are marked with the invisible digital sign SynthID - this is how Google labels content created by artificial intelligence.