Demis demo: Lyria 3 Pro music
Demis Hassabis shared a demo of Lyria 3 Pro that generates up to three‑minute, structured music pieces—complete with intros, verses and bridges—and is available through Gemini App and Google AI Studio. The demo underscores how generative audio is moving from short loops to compositional structure. (x.com)
Google published its Lyria 3 Pro announcement on March 25, 2026 via an official blog post and a Workspace Updates entry. (blog.google) (blog.google) Longer-generation access in the Gemini app is being rolled out starting with paid subscribers, according to Google’s product notes. (blog.google) (blog.google) Developers and businesses can call the model in Vertex AI as a public preview using the documented model identifiers (including lyria-3-pro-preview) and receive downloadable WAV results from the API. (cloud.google.com) (docs.cloud.google.com) Google DeepMind’s SynthID system embeds an inaudible, robust watermark in Lyria-generated audio to enable detection of AI-origin content. (deepmind.google) (deepmind.google) Workspace Updates lists Lyria 3 Pro availability in multiple languages (English, Japanese, Korean, Hindi, Spanish, Portuguese, German, French) and notes the Gemini music-generation feature requires users to be 18 or older. (workspaceupdates.googleblog.com) (workspaceupdates.googleblog.com) Google is integrating Lyria 3 Pro across creator surfaces including Google Vids and its ProducerAI tooling, with initial rollouts targeting Workspace customers and Google AI Pro/Ultra subscribers. (blog.google) (blog.google) Google’s developer announcements specify a paid-preview launch model for Lyria 3 (pay-per-request for API use) while also pointing to a prompt guide and examples for structured music generation in the Vertex AI documentation. (aiproductivity.ai) (aiproductivity.ai)
