Google this week announced the launch of the beta version of Lyria 3 in the Gemini app, marking the company’s next step in offering creative expression tools.
Google DeepMind’s new generative music model allows users to create custom tracks from text commands or media files such as photos and videos. The tool is designed to transform ideas — from comedic concepts to nostalgic memories — into high-quality compositions in a matter of seconds.
News
Lyria 3 presents significant advances in relation to , standing out in three main pillars: lyrical automation, which eliminates the need for the user to provide their own lyrics; greater creative control over technical elements such as style, vocals and tempo; and the ability to generate tracks with greater musical complexity and realism.
Features and integration
The versatility of the new model allows for different creative approaches. Through the text-to-track function, the system interprets specific genres, moods and contexts to generate instrumentals or songs with original lyrics. Additionally, the photo and video to track modality uses visual content sent by the user to compose tracks that match the atmosphere of the image.
Gemini app productions have the following characteristics:
- Duration: 30 second tracks.
- Visual identity: custom covers generated by the Nano Banana model.
- Sharing: direct download options and social media links.
In addition to the app, Lyria 3 is being integrated into YouTube’s Dream Track. Initially available in the US and now expanding to other markets, the model will help content creators take their Shorts to the next level by customizing exclusive backing tracks and lyrical verses.
Security and responsible development
Accompanying the launch, Google introduced new audio verification features. All generated tracks are flagged with SynthID, an imperceptible watermark designed to identify content created by the company’s artificial intelligence.
Gemini is now also capable of analyzing audio files uploaded by users to verify, through SynthID and logical reasoning, whether the work was produced by
Regarding ethics and copyright, the company reinforced its commitment to responsible development.
Lyria 3 was trained under agreements with music industry partners and designed to generate original expressions while avoiding direct imitation of existing artists. If an artist is mentioned in a command, the system uses the reference only as style inspiration or general “vibe”.
Currently, Lyria 3 is available to users over 18 years old in eight languages, including Portuguese. Although access is broad, subscribers to the Google AI Plus, Pro and Ultra plans have higher usage limits.
*Published by André Nicolau, from CNN Brasil
