Google Lyria 3 Pro: AI now creates full three-minute tracks

Google has unveiled the updated music generation model, Lyria 3 Pro. Released a month after Lyria 3, the new version allows users to create tracks up to three minutes long, compared to just 30 seconds with the previous model.

Users now have greater control over the composition’s structure. Intros, verses, choruses, and transitions can be specified separately, as the model better understands the sequence and logic of musical elements.

Where the model can be used

Lyria 3 Pro is available through the Gemini app, similar to the previous version, but only for paid subscribers. The model is also integrated into the Google Vids video editor and the ProducerAI music creation tool. Corporate Google services, including Vertex AI (in public testing), Gemini API, and AI Studio, have also received Lyria 3 Pro functionalities.

How the model was trained and AI track identification

For training, Google used partner data, authorized YouTube materials, and its own internal resources. Google emphasizes that the model does not copy the style of specific artists but can take “broad inspiration” from performers mentioned in the prompt.

All tracks created with Lyria 3 and Lyria 3 Pro carry a SynthID watermark, indicating that the composition was generated with the assistance of artificial intelligence.