Google has launched what it says is its ‘most advanced’ generative AI music model yet, Lyria 3, within its Gemini chatbot app — allowing users to create 30-second tracks from text prompts or images.
Lyria 3 was developed by Google DeepMind, the tech giant’s AI research division, and is the latest iteration of its generative music technology.
The feature, which began rolling out on Wednesday morning (February 18), enables users aged 18 and over to describe a genre, mood, or scenario and receive an AI-generated track complete with vocals and lyrics. Users can also upload photos or videos and have Gemini compose an accompanying soundtrack.
Google is positioning the tool for casual use cases such as custom birthday cards or “a ballad from your pet’s POV.” Lyria 3 is available in English, German, Spanish, French, Hindi, Japanese, Korean and Portuguese.
Google has not specified how Lyria 3 was trained, but said in a blog post on Wednesday (February 18) it has sought to “develop this technology responsibly in collaboration with the music community” and has “been very mindful of copyright and partner agreements” in training the model.
MBW understands that to mean the training for Lyria 3 uses music that YouTube and its parent company, Google, “have the right to use” under their “terms of service, partner agreements, and applicable law”.
Billboard reported in January 2024 that Google had previously trained its AI music models on copyrighted recordings before approaching rights holders for licensing deals.
Google said Lyria 3 represents an improvement on its previous Lyria models, offering automatically generated lyrics, greater creative control over style and tempo, and “more realistic and musically complex tracks.”
The new model builds on Lyria 2, which MBW reported on in September, and which powers YouTube’s “Speech to Song” tool for converting dialogue into musical soundtracks.
The company framed the feature as a tool for personal creative expression rather than professional music production. “The goal of these tracks isn’t to create a musical masterpiece, but rather to give you a fun, unique way to express yourself,” wrote Joël Yawili, Senior Product Manager at Gemini App, and Myriam Hamed Torres, Senior Product Manager at Google DeepMind, in a blog post announcing the launch.
Lyria 3 is also being integrated into YouTube’s Dream Track — the platform’s AI soundtrack generator for Shorts creators, first launched in 2023 — which is expanding beyond its initial US availability to creators in other countries.
Dream Track originally launched with AI voice clones of participating artists including Charlie Puth, T-Pain, and Alec Benjamin. Lyria 3 focuses on original vocal and instrumental generation rather than artist likeness.
The launch follows YouTube’s rollout of a separate AI music tool, “Music Assistant,” within its Creator Music marketplace earlier this year. That feature, which MBW reported on in April last year, allows YouTube Partner Program members to generate copyright-free instrumental backing tracks via text prompts.
Google said that Lyria 3 is “designed for original expression, not for mimicking existing artists.” If a user’s prompt names a specific artist, the company said, “Gemini will take this as broad creative inspiration and create a track that shares a similar style or mood.”
The company added that it has “filters in place to check outputs against existing content” and that users can report content that may violate their rights.
All tracks generated in Gemini are embedded with SynthID, Google’s watermarking technology for identifying AI-generated content. The company also announced that users can now upload audio files to Gemini to check whether they were generated using Google AI.
The launch comes after Universal Music Group and YouTube struck a new licensing deal in October that included what UMG CEO Sir Lucian Grainge called “really important guardrails and protection for our artists and writers around gen AI content.”
Music Business Worldwide