Google's Lyria 3 Pro AI model now creates music up to 3 minutes long, offering deeper structure understanding for compositions, vlogs, and podcasts. Available to Gemini subscribers.
© A. Krivonosov
Google has expanded the capabilities of its Gemini AI by adding an enhanced music model called Lyria 3 Pro. This new version can now create full compositions lasting up to three minutes, a significant increase from the previous 30-second limit.
Lyria 3 Pro offers deeper understanding of musical structure. It can account for compositional elements like intros, verses, choruses, and transitions, and it follows complex user requests more precisely. Users can now provide specific instructions for individual song sections rather than just describing a track's overall style, leading to more accurate and thoughtful results.
The update makes music generation more flexible: the model works well for creating original compositions as well as practical tasks like producing background music for vlogs, podcasts, or educational videos. With improved track-building logic, the AI can create more cohesive and dynamic works featuring smooth transitions between sections.
Lyria 3 Pro is already available to paid Gemini subscribers, though daily track creation limits apply. Google AI Plus users can generate up to 10 compositions, Pro users up to 20, and Ultra subscribers up to 50 tracks per day. Beyond the Gemini app, the model is also accessible on other platforms including Vertex AI, Google AI Studio, Gemini API, Google Vids, and ProducerAI.
Google emphasizes that the Lyria family of models was developed with input from music industry professionals and trained on legally available data. The AI doesn't copy specific artists even when users mention them in requests—such references serve only as inspiration. All generated tracks feature a SynthID watermark that helps distinguish AI content from original works.