Google has unveiled its latest AI music model, Lyria 3, in beta as part of the Gemini app, with the announcement made on Wednesday, February 18. This innovative tool allows users to generate up to 30-second songs by inputting text prompts and uploading photos or videos, all while choosing from a range of popular musical styles. Each generated track is complemented by AI-created cover art from Google’s image model, Nano Banana.
The press release emphasizes that Lyria’s primary aim “isn’t to create a musical masterpiece, but rather to give you a fun, unique way to express yourself.” This third version of the Lyria model introduces three significant enhancements. Previously, users were required to supply their own lyrics and had limited creative control over aspects such as style, vocals, and tempo. Additionally, earlier iterations produced tracks that were less musically intricate.
Importantly, Lyria 3 has been “designed for original expression, not mimicking existing artists,” according to the release. It contains “filters in place to check outputs against existing content,” though the company acknowledges that this approach “might not be foolproof.” Users can report any tracks that potentially violate copyright rights.
In a response to inquiries regarding the model’s training data, a representative clarified that Lyria is conscientious about copyright and partner agreements. The model is trained exclusively on music that YouTube and Google have the right to use under their terms of service, partner agreements, and relevant laws. Consequently, Lyria is restricted from creating songs based on specific artists, songs, books, and other protected materials.
However, through testing conducted by Billboard, it became evident that Lyria possesses an understanding of the characteristics of these restricted keywords, allowing it to generate songs that encapsulate their essence without directly imitating them. For instance, when prompted to create “a love song as if you are Heathcliff from Wuthering Heights singing to Cathy in the style of Kate Bush,” Lyria responded with a track that captured “the fierce, sweeping drama” of the request. The generated song was described as “dramatic ethereal art-pop…steeped in gothic romanticism…featur[ing]…atmospheric synthesizers,” and even included references to Heathcliff and “the lonely moors,” showcasing a nuanced understanding of both Bush’s musical style and the thematic elements of Emily Brontë’s novel.
Users have the option to download their generated songs as MP3 files or as videos that combine the music with the corresponding cover art. All tracks produced within the Gemini app are embedded with SynthID, an invisible audio watermark designed to identify content generated by Google’s AI.
As the integration of AI into creative fields continues to evolve, Lyria 3 stands out as an example of how technology can democratize artistic expression, allowing users to explore their creativity without needing advanced musical skills. The intersection of technology and art raises broader questions about the implications of AI-generated content on the music industry, particularly regarding copyright, creativity, and originality.
See also
AI Characters Redefine Fandom Dynamics as Community-Driven Canon Emerges
Anthropic Study Reveals Rising AI Agent Autonomy with 40-Minute Sessions in Coding
Barclays Highlights Physical AI’s Growth Potential as Market Approaches Inflection Point
Mistral AI and Ericsson Partner to Revolutionize Telecom with Advanced AI Solutions
Cohere Co-Founder Nick Frosst Advocates for Stronger Canadian AI Amid Rapid Growth
