Google unveils Lyria 3 Pro, its latest AI music generation model

Google on Wednesday introduced Lyria 3 Pro, its latest music-generation model, just a month after the release of Lyria 3. The upgraded model enables users to generate music tracks up to 3 minutes long, a notable increase over the 30-second limit of the earlier Lyria 3 version.

Alongside its extended track length, the company said the Lyria 3 Pro delivers improved creative control and deeper customisation. Users can now define specific elements of a composition in prompts, including intros, verses, choruses, and bridges, as the model demonstrates a stronger understanding of song structure than its predecessor.

Google previously integrated music-generation features into the Gemini app with the launch of Lyria 3. The new Pro version is also being introduced within the Gemini app, though access will be limited to paid subscribers.

The company is further expanding the model’s reach by bringing Lyria 3 Pro to its video-editing platform, Google Vids, and to ProducerAI, a generative AI-powered music production tool that Google acquired last month.

In addition, Google is extending music generation capabilities to enterprise users. Lyria 3 Pro is being added to Vertex AI in public preview, as well as to the Gemini API and AI Studio, allowing developers and businesses to integrate music creation into their workflows.

Google noted that the model was trained on data from its partners, as well as permissible datasets from YouTube and other Google services. The company stated that the model does not replicate any specific artist’s work. However, when users reference an artist in prompts, the system draws only “broad inspiration” from that style when generating music.

To ensure transparency, all tracks created with Lyria 3 and Lyria 3 Pro are labelled with SynthID, indicating that the content was generated using AI.

The launch comes amid broader industry efforts to address challenges in AI-generated music. Earlier this week, Spotify introduced tools allowing artists to review tracks published under their names to prevent misattribution, while Deezer has developed technology to help platforms detect AI-generated songs.