Google's Gemini AI Makes Music
Google's Gemini AI can now generate music through its new Lyria 3 feature, allowing users to create 30-second original tracks by typing descriptions or uploading photos and videos. The feature is accessible at gemini.google.com under 'create music' according to the Rich On Tech podcast Episode 162 from February 21. This offers content creators and marketers a fast way to generate unique soundtracks for presentations and social media.
- Lyria 3 was developed by Google's AI research division, DeepMind, and improves upon previous models like Lyria 2, which powered a "Speech to Song" tool for YouTube. - A key advancement in Lyria 3 is its ability to generate its own lyrics based on a user's prompt, in addition to offering more granular control over musical elements like style and tempo. - To identify AI-generated content, all tracks created with Lyria 3 are imperceptibly watermarked using a technology called SynthID. - While Google has not specified the exact training data, it states the model was trained on music it has the right to use via partner agreements and terms of service, which is understood to include licensed content from YouTube. - The feature began rolling out on February 18, 2026, to users aged 18 and over in languages including English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese. - Beyond the Gemini app, the Lyria model is also being integrated into other Google products, such as Dream Track, an AI soundtrack generator for YouTube Shorts creators. - This launch positions Google's Gemini in direct competition with other established AI music generation services like Suno and Udio. - In addition to creating tracks from prompts, the tool can also generate custom cover art for the music it produces using a model called Nano Banana.
