Google is adding an AI-based music generation feature to its Gemini app, TechCrunch reported on Feb. 18.

The Gemini app uses DeepMind’s Lyria 3 model to automatically generate music and lyrics in a style users want. For example, if a user asks for a “comic R&B slow jam about socks finding their match,” it generates a 30-second track and even cover art. Users can also upload photos or videos to create music that matches the mood.

Lyria 3 can generate more realistic and complex music than the previous generation, and users can adjust style, vocals and tempo. Google said the update will also apply Lyria 3 to its Dream Track feature for YouTube creators. The feature was previously available only in the United States, but is now expanding globally.

Imitating specific artists is not possible. Google said Lyria 3 is “a tool for creative expression rather than mimicking existing artists,” adding that if a specific artist is entered in a prompt it generates a track with a similar style or mood, but a filter is also applied to compare it with existing content.

All AI-generated music will have a SynthID watermark, and the Gemini app will also add a feature to identify whether content is AI-generated. The music generation feature supports eight languages: English, German, Spanish, French, Hindi, Japanese, Korean and Portuguese. It will be rolled out sequentially to users aged 18 and older worldwide.