Google Launches Lyria 3 Pro: Longer Tracks, Deeper Control, and Enterprise Access

Just one month after releasing Lyria 3, Google is already stepping up its AI music ambitions with the launch of Lyria 3 Pro — a more capable tier of its music generation model that pushes beyond the limits of what the original release could do.

Longer Tracks, Smarter Structure

The most immediate upgrade is track length. Where Lyria 3 capped generated audio at 30 seconds, Lyria 3 Pro extends that ceiling to three full minutes — a meaningful shift for anyone attempting to produce a complete, usable piece of music rather than a clip or a demo fragment.

Beyond duration, the Pro model brings stronger compositional intelligence. Users can now specify individual structural elements within a prompt — intros, verses, choruses, bridges — and Lyria 3 Pro will interpret and honor that architecture during generation. The model understands track structure in a way its predecessor did not, offering a level of creative direction that moves AI music generation closer to genuine production tooling.

Where It’s Available

Google is deploying Lyria 3 Pro across several surfaces simultaneously. In the Gemini app, where Lyria 3 first brought music generation to consumers, the Pro model is rolling out now — though access is gated to paid subscribers. Google Vids, the company’s AI-assisted video editing tool, is also gaining Lyria 3 Pro integration, giving video creators an embedded music generation layer directly inside their editing workflow.

The acquisition context matters here too: ProducerAI, a generative AI music production platform that Google acquired last month, is receiving Lyria 3 Pro as its core engine — a clear signal that Google intends ProducerAI to become a serious destination for music creators working with AI tools.

On the enterprise and developer side, Lyria 3 Pro is entering Vertex AI in public preview, alongside availability through the Gemini API and AI Studio, opening the model to developers building music-integrated applications at scale.

Training, Attribution, and Watermarking

Google was deliberate in addressing how the model was built and how it handles artist identity. Lyria 3 Pro was trained on data from Google’s partners, permissible YouTube content, and Google’s own data. The company states the model does not replicate or mimic any specific artist’s style — but acknowledges that if a user names an artist in a prompt, the model draws “broad inspiration” from that artist’s aesthetic without attempting direct imitation.

Every track generated by either Lyria 3 or Lyria 3 Pro is watermarked with SynthID, Google DeepMind’s tool for identifying AI-generated audio — a measure designed to keep the provenance of machine-made music traceable.

A Platform Moment for AI Music

The launch arrives as the broader streaming and music industry grapples with how to handle the volume of AI-generated content entering distribution pipelines. Earlier this week, Spotify rolled out new tools giving artists the ability to audit songs released under their name, responding to cases where AI-generated tracks have been misattributed to real musicians. Separately, Deezer announced detection infrastructure designed to identify AI-generated music across streaming platforms — tools it has offered to share with the wider industry.

Google’s Lyria 3 Pro positions the company at the center of that evolving landscape: not just as a model provider, but as a platform building the infrastructure — generation, distribution, and watermarking — around AI music end to end.