Google launches Lyria 3.5 music generation in Gemini App with SynthID watermark
Google expanded Lyria 3.5 music generation to the Gemini App and API on September 4, giving 950M monthly active users access to full-length songs with imperceptible SynthID watermarks.
Google expanded its Lyria 3.5 music generation model to the Gemini App, the Gemini API, Google AI Studio, and Google Vids on September 4, giving the app's 950 million monthly active users the ability to generate full-length songs with vocals and instrumentals from text prompts. The model outputs 44.1 kHz stereo audio with track lengths ranging from 30 seconds to roughly three minutes.
The Gemini App version of Lyria 3.5 adds genre selection, vocal/instrumental switching, templates for background music and custom birthday tracks, and the ability to upload up to 10 images as creative prompts. Every generated track ships with an imperceptible SynthID audio watermark, satisfying provider-side marking obligations under the EU AI Act. Lyria 3 Clip, the 30-second preview model, ships under the ID lyria-3-clip-preview; Lyria 3.5 ships under lyria-3.5.
Developers access both models through the Interactions API, with MP3 as the default output and WAV available for Lyria 3.5. Safety filters block requests for specific artist voices and copyrighted lyrics. Generation is single-turn and non-deterministic.