Google adds Lyria 3 to Gemini, creates 30-second music from text and images

Google has begun deploying the Lyria 3 music generation model within the Gemini app (beta version), allowing users over 18 to create 30-second songs based solely on text descriptions or uploaded images. The system automatically generates full lyrics, background music, and cover art using AI.

Lyria 3, developed by Google DeepMind, produces coherent music quality that closely follows the description, but is currently limited to niche genres and a maximum length of 30 seconds. Google positions this as a tool for creating “shareable moments” rather than full commercial songs.

Compared to competitors like Suno and Udio—which support multi-minute songs with complete structures and multiple control parameters—Lyria 3 remains more basic for in-depth music production.

Regarding copyright, competitors have previously been sued by the Recording Industry Association of America; Udio has reached agreements with Warner Music and aims for a fully licensed platform. Google states that Lyria 3 was trained with partner agreements and avoids directly mimicking specific artists.

All music generated within Gemini is marked with a SynthID audio watermark and includes source verification tools. Streaming platforms like Deezer have also implemented AI music detection systems. Lyria 3 is simultaneously expanding the Dream Track feature on YouTube for Shorts creators worldwide.