3, in beta. It enables users to create 30-second tracks using text prompts, images or videos.
Lyria 3 is available in the Gemini app for users aged 18 and above in English, German, Spanish, French, Hindi, Japanese, Korean and Portuguese. The feature is rolling out on desktop and will expand to mobile in the coming days. Google AI Plus, Pro and Ultra subscribers will have higher usage limits.
Currently in the US and expanding, the feature is also available to creators on YouTube’s Dream Track and lets creators make custom soundtracks for Shorts.
The company said Lyria 3 allows users to describe an idea or upload media, which is then converted into a track within seconds by Gemini.
The model generates lyrics automatically based on prompts and offers controls over style, vocals and tempo. Google added that Lyria 3 delivers more complex and realistic compositions compared to earlier versions.
Users can create tracks by describing a genre, mood or memory. For example, a prompt such as creating “a track for my mother about the great times we had as kids” in Afrobeat style will generate a corresponding song. The app also allows users to upload photos or videos and generate tracks that reflect the content.
Each 30-second track includes custom cover art generated by Nano Banana and can be downloaded or shared via link. “The goal of these tracks isn’t to create a musical masterpiece, but rather to give you a fun, unique way to express yourself,” the company said.
Google said all tracks generated in the Gemini app are embedded with SynthID, its watermarking technology for AI-generated content. The company has expanded its verification tools to include audio, allowing users to upload files and check whether they were created using Google AI.
“Music generation with Lyria 3 is designed for original expression, not for mimicking existing artists,” it added.
If users reference a specific artist, Gemini will interpret it as inspiration and generate music in a similar style rather than replicate existing work. Filters are in place to compare outputs against existing content, and users can report potential violations.
ALSO READ: Anthropic Launches Claude Sonnet 4.6 with 1M Token Context Window
