Google said on the 19th that it has added a music generation feature to its Generative AI app Gemini. When a one-line text prompt (command) or an image is entered, Gemini creates an audio track about 30 seconds long along with lyrics.
Google said it added to Gemini a music generation feature based on Lyria 3, Google DeepMind's state-of-the-art music generation model, on the same day. The feature is currently in beta, with an official release planned later.
When a specific genre or mood is entered, Gemini generates lyrics and a melody based on it and completes them as a single track. For example, if you request "a comedic R&B slow jam about a sock that found its match," Lyria 3 generates an audio track within seconds. If you upload a photo or video, it also creates music that matches the mood of that content.
Along with this, Google's image generation tool Nano Banana also creates cover art that suits the music. The finished song can be downloaded or shared with acquaintances or friends via a link.
Google said, "It is not permitted to directly imitate a specific artist." However, it noted that if an artist's name is mentioned in a prompt, Gemini generates music in a style or mood similar to that artist.
All audio generated with Lyria 3 includes SynthID, a digital watermark indicating that the content was created by AI. A feature to determine whether music uploaded to Gemini was generated by AI will also be added.
Google also equipped Dream Track, an AI-based music creation tool available to YouTube creators, with Lyria 3. It is currently offered only in the United States, but will be rolled out to other countries in stages.
The music generation feature supports English, German, Spanish, French, Hindi, Japanese, Korean and Portuguese, and is available to users 18 and older. It will be applied in stages starting with the desktop version and expanded to the mobile app within a few days.