Google rolls out Lyria 3 music model inside Gemini

Google has begun rolling out its Lyria 3 AI music model in the Gemini app (beta), allowing users aged 18+ to generate full 30-second songs from text prompts or uploaded images, including vocals, instrumentals, and AI cover art.

Built by Google DeepMind, Lyria 3 produces coherent, prompt-aligned tracks within seconds, but currently struggles with niche genres and caps output at 30 seconds. Google positions the tool for quick, shareable music snippets rather than fully polished commercial songs.

Competitors such as Suno and Udio still lead in long-form generation, advanced controls, and structured multi-minute compositions.

All generated tracks include Google’s SynthID audio watermark and built-in verification tools. The model is available on desktop now, with mobile rollout following, and is also expanding Dream Track creation on YouTube Shorts globally.