Google Lyria 3 Pro: New AI Music Generation Model
▼ Summary
– Google released Lyria 3 Pro, a new music generation model that creates tracks up to three minutes long, a significant increase from the previous 30-second limit.
– The model offers improved creative control, allowing users to specify structural elements like intros and choruses due to its better understanding of track composition.
– Access to Lyria 3 Pro is limited to paid subscribers in the Gemini app and is also being integrated into Google Vids and the acquired ProducerAI tool.
– The model is being added to Google’s enterprise platforms, including Vertex AI, the Gemini API, and AI Studio for business use.
– All music generated by Lyria models is watermarked with SynthID to denote AI creation, and the model is trained on partner and permissible YouTube/Google data without directly mimicking artists.
Google has unveiled its latest AI music generation model, Lyria 3 Pro, just weeks after introducing its predecessor. This new release significantly expands creative possibilities by enabling users to produce full-length tracks of up to three minutes, a major leap from the previous 30-second limit. The upgrade is designed to provide musicians and creators with far greater creative control and customization over their compositions.
Beyond simply extending track duration, Lyria 3 Pro offers enhanced structural understanding. Users can now direct the model by specifying distinct song sections like intros, verses, choruses, and bridges within their prompts. This allows for more sophisticated and coherent musical arrangements that follow traditional songwriting formats. The model is being integrated into the Gemini app, though access to the Pro version will be reserved for paying subscribers. It is also rolling out to Google’s other creative platforms, including the Google Vids video editor and the recently acquired ProducerAI music production tool.
The expansion of this technology is not limited to consumer applications. Google is also bringing music generation capability to its enterprise ecosystem. Lyria 3 Pro will be available through Vertex AI, currently in public preview, as well as via the Gemini API and AI Studio, offering businesses new tools for content creation.
In addressing the critical issue of training data and originality, Google stated that Lyria 3 Pro was developed using data from its partners and permissible sources from YouTube and Google. The company emphasized that the model is not designed to mimic specific artists. However, if a user’s prompt includes an artist’s name, the system will take broad inspiration from that creator’s style without producing a direct imitation. To ensure transparency, all music generated by both Lyria 3 and Lyria 3 Pro models will be watermarked with SynthID, a digital identifier that labels the track as AI-generated.
This move by Google arrives amid a broader industry focus on managing AI-generated content. Earlier this week, Spotify introduced new tools allowing artists to review songs released under their name to prevent misattribution by AI creators. Similarly, streaming service Deezer has launched technology to help any platform identify AI-made music, highlighting the growing push for clarity and authenticity in the digital music space.
(Source: TechCrunch)
