Google DeepMind Drops Lyria 3, Its Most Capable Generative Music Model

Lyria 3 can turn text prompts and photos into full tracks with realistic vocals, lyrics, and granular control over tempo and style — and it's already live in Gemini as a beta.

Google DeepMind released Lyria 3, its latest generative music model, with immediate availability as a beta feature inside the Gemini app. As @GoogleDeepMind detailed, the model produces 48kHz stereo audio with "crystal-clear" fidelity and supports granular controls for tempo, lyrics, instrumentation, and vocal style. In a separate post, the team highlighted that Lyria 3 can also generate music from photos — a multimodal input capability that previous music models lacked.

Unlock the full briefing

Get every story in today's briefing, the full archive, and the daily AI intelligence brief.

All stories today

Full archive

Daily brief

Cancel anytime. Payments powered by Stripe.