Google integrated a new music tool based on the Lyria 3 model into its Gemini neural network. It can transform a regular text query into a full-fledged music track with vocals — simply describe the mood, genre, or situation.
The new version of Lyria 3 operates in beta mode directly within the Gemini app, without additional services or plugins. Unlike previous versions, it can autonomously create song lyrics and music. Users can also control the tempo, style, and vocal type, and the sound has become more complex and realistic.
The system works simply: you can enter a query or upload a photo or video. For example, if you describe the desired composition, Gemini will generate lyrics and music, resulting in a 30-second clip with an automatically created cover. The ready track can be downloaded or shared via a link.
If you upload an image, the model analyzes its mood and creates corresponding music. However, it is not possible to exactly copy the style of a specific artist — the system uses it only as a general reference. All generated compositions are marked with an invisible SynthID watermark, which allows identifying that the track was created by artificial intelligence.
The feature is available to users over 18 years old in the desktop version, and mobile devices will receive it in the coming days. Several languages are supported. Lyria 3 also appeared in the Dream Track tool on YouTube, which helps creators add AI music to short videos.
Comments