Google said it is adding a music-generation feature to its Gemini app, powered by Lyria 3, a model developed by DeepMind. The feature is launching in beta and allows users to generate 30-second music tracks with lyrics by describing a song concept in text. Users can also upload photos or videos, prompting Gemini to create music that matches the mood of the media.
Lyria 3 improves on earlier versions by producing more complex and realistic compositions. Users can adjust elements such as musical style, vocals, and tempo. Generated tracks include cover art created by Google’s Nano Banana image system.
Google is also expanding access to Lyria 3 through Dream Track on YouTube, enabling creators worldwide to generate AI-assisted music. The tool was previously limited to U.S. creators. Google said the system is designed for original output and does not directly mimic specific artists, even when names are included as creative references.
All music generated with Lyria 3 will include a SynthID watermark to identify AI-created content. The feature is rolling out globally to users aged 18 and older, supporting multiple major languages.
The launch follows Google’s recent release of Gemini 3 Deep Think, a specialized reasoning mode for scientific and engineering tasks, underscoring the company’s broader push to position Gemini as a core platform for advanced AI capabilities.
