Google has switched on Lyria 3.5, its music-generation model, for all Gemini users in early September, turning a text prompt into a complete song.
The company is rolling out the model globally on the web and in the mobile Gemini app, and also offers it through the Gemini API, Google AI Studio, Google Flow Music and Google Vids. Google calls Lyria 3.5 its best-sounding music model yet, with more expressive vocals and richer arrangements than its predecessor.
Full songs, not just clips
The headline change is length. Where the older Lyria 3 Clip produced fixed 30-second snippets, Lyria 3.5 generates multi-minute tracks with verses, choruses and bridges – up to roughly three minutes, according to Google. Inside the Gemini app, guided controls walk users through the process:
- Pick a genre or style, or describe one in your own words
- Choose between vocals or an instrumental version
- Start from templates for background music, jingles, ringtones or birthday songs
Every generated file carries a SynthID watermark that Google says is inaudible to humans and flags the audio as AI-made. On training data, Google states the Lyria model was trained only on licensed content but discloses no specifics. The company did not announce API pricing at launch.
By putting song generation inside a mainstream app used by hundreds of millions, Google is likely to reignite debate over copyright and how musicians are compensated when AI mimics their craft.
Sources: The Decoder · Unite.AI
