Google’s Lyria 3 Pro Generates Full 3‑Minute AI Songs

Google has rolled out Lyria 3 Pro, a significant update to its music‑generation model that extends output length and adds finer control over song structure. Where the initial release capped samples at around 30 seconds, the Pro version can now produce full tracks up to three minutes long, making the outputs more suitable for demos and longer musical ideas.

Beyond length, Lyria 3 Pro accepts more detailed prompts. Users can request intros, verses, choruses and bridges, and the model is designed to better understand musical composition and transitions. That extra granularity should help creators experiment with arrangements and explore stylistic shifts inside a single generated piece.

The tool is live for paid Gemini subscribers and enterprise customers on Vertex AI. Developers can tap Lyria 3 Pro through the Gemini API and Google AI Studio, and Google plans to integrate the tech with its Google Vids video‑generation platform to streamline audiovisual workflows.

Google emphasizes responsible training and rights management: the company says Lyria 3 Pro was trained only on materials it has rights to, and every output is embedded with SynthID, Google’s watermark for identifying AI‑generated media. That approach aims to make provenance clearer as AI‑created music proliferates.

Still, the update arrives amid a flood of AI music online. Estimates suggest tens of thousands of AI‑generated tracks are uploaded to streaming services daily; Spotify reportedly removed millions of spammy tracks last year. Lyria 3 Pro’s longer, more customizable songs will likely intensify debates about discovery, platform moderation and the role of AI in music creation.