DeepMind unveils Lyria 3 Pro for high-end AI music creation: Here’s how it works

Lyria 3 and Lyria 3 Pro are now available for developers through the Gemini API and Google AI studio. The models combine musical awareness with structural coherence, allowing for consistent, high-fidelity compositions. The premier model, Lyria 3 Pro is designed for studio-quality, full-length songs up to approximately three minutes long. Meanwhile, Lyria 3 is optimized for speed and high-volume requests, generating 3–second clips ideal for social media loops, and prototyping. Both variants support realistic, expressive vocals in multiple languages and across diverse genres.

Google AI studio

A new dedicated workspace for experimenting with two modes:

• Text Mode

Generate music using natural language description, including specific parameters like tempo or key.

• Composer mode

Build songs section-by-section for granular control. This mode also supports advanced features like image-to-music generation. Paid Gemini users can start creating up to 184-second tracks or quick 30-second clips today through the app, with developer access via APIs starting at $0.08 per song. Early tests include DeepMind CEO Demis Hassabis’s late-night instrumental and a boy-band remix of poetry-all of which are protected by content filters. To ensure transparency, every track includes a SynthID digital framework that remains detectable even if the audio is modified. Furthermore, the technology was developed to ensure the AI serves as a tool for human creativity rather than a replacement.