On Wednesday local time in the United States, Google announced the launch of a new Lyria 3 Pro music generation model, which is another upgrade after the official release of Lyria 3 last month. Compared to the previous Lyria 3, which generated music clips of up to 30 seconds, the new version supports the generation of tracks up to 3 minutes in length, and has comprehensively enhanced creative control and customization capabilities.
Google said that Lyria 3 Pro not only extends the time for generating music, but also allows users to more precisely specify the structure of the work in the prompt words, including different parts such as the intro, verse, chorus, and bridge. The model's ability to understand the structure of the complete track has been significantly improved compared to the previous generation. Users can use more detailed text descriptions to guide the model to generate music works that are closer to expectations.
In terms of product layout, Google has previously introduced the music generation function into the Gemini application through Lyria 3, and this time the Pro version will also be launched on Gemini, but it will only be open to paying subscribers. At the same time, Lyria 3 Pro will also be integrated into Google’s video editing application Google Vids, as well as the generative AI music production tool ProducerAI acquired by Google last month, further enriching its AI product matrix in the field of content creation.
In terms of enterprise-level applications, Google will introduce music generation capabilities based on Lyria 3 Pro through Vertex AI (currently in public preview), Gemini API and AI Studio, providing developers and enterprises with music generation capabilities that can be integrated into their own products and services. This means that everyone from individual creators to large institutions can call this model on Google Cloud and related development platforms for scenarios such as video soundtracks, game sound effects, or customized sound brands.
On the issue of training data and copyright compliance, Google emphasized that Lyria 3 Pro uses data from partners, as well as "licensable data" from YouTube and Google. The company also emphasized that the model will not "imitate" specific artists, but if the user names an artist in the prompt, the model will get "broad inspiration" from the artist to create music of the corresponding style.
To identify AI-generated content, Google said that all audio tracks generated through Lyria 3 and Lyria 3 Pro will be embedded with a SynthID tag, indicating that the work was created with the participation of artificial intelligence. At a time when regulatory and industry concerns about AI music are heating up, this marking mechanism will help subsequent platforms provide technical basis when identifying and managing AI content.
In the broader industry context, streaming platforms are also accelerating their response to AI music. Earlier this week, Spotify launched a new tool that allows artists to review songs released under their own names to prevent so-called "AI spam" from being mistakenly attributed to real musicians. Another music platform, Deezer, has launched related tools to help any streaming service identify AI-generated music, allowing the platform to adopt a clearer strategy when facing this type of content.