Blog

On Wednesday, Google revealed the integration of a music creation tool into its Gemini application. This new feature, currently in beta, is driven by DeepMind’s Lyria 3 music-generation model.

Users can activate this capability by describing their desired song, prompting the app to produce both a musical track and its lyrics. For example, a request to Gemini for a “comical R&B slow jam about a sock finding its match” would result in a 30-second piece of music, complete with accompanying cover art designed by Nano Banana.

Google also stated that the AI-driven tool can generate music that reflects the mood of an uploaded photo or video.

According to the company, Lyria 3 represents an advancement over earlier models, enabling the creation of more sophisticated and lifelike music. Additionally, users gain control over various aspects such as style, vocal arrangements, and tempo.

In addition to its deployment in the Gemini app, Google is extending Lyria 3’s capabilities to YouTube creators via the Dream Track feature, a mechanism for generating AI-powered music. Previously, this option was exclusively accessible to U.S. YouTube creators, but this launch marks its global expansion.

Google clarified that direct imitation of artists is not possible; however, including an artist’s name in a prompt will guide Gemini to produce a track evoking a comparable style or mood. (It remains uncertain whether this generation method might inadvertently facilitate the deciphering of specific artists’ musical styles.)

“The Lyria 3 music generation system is intended for unique creations, not for copying established artists. Should a user’s prompt mention a particular artist, Gemini will interpret this as general creative guidance, crafting a track with a similar feel or genre. We also employ filters to compare generated content with existing works,” stated the company in a blog post.

Google highlighted that every song produced using the Lyria 3 model will incorporate a SynthID watermark, serving to mark it as AI-generated. The company further mentioned that it is integrating features into Gemini to identify AI-generated music via SynthID. This will allow users to upload tracks and inquire whether Gemini recognizes them as AI-produced.

The music generation feature is being progressively released worldwide to all Gemini users aged 18 and above, supporting multiple languages including English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese.

The emergence of AI-generated music has elicited varied reactions from both creators and audiences. While entities such as YouTube and Spotify are embracing AI, entering agreements with music labels to profit from AI-created tracks, AI development firms and tool providers are simultaneously encountering legal challenges from the music sector concerning the copyright of their training data. Furthermore, platforms like Deezer have introduced mechanisms to identify AI-generated music, aiming to prevent illicit streaming of such content.