Particle.news
Download on the App Store

Google Brings DeepMind’s Lyria 3 to Gemini for 30‑Second AI Music

The beta centers on playful, short‑form creation with SynthID watermarking plus guardrails to curb direct artist mimicry.

Overview

  • Lyria 3 in the Gemini app now generates 30‑second songs from text prompts or uploaded photos and videos, with auto‑written lyrics, vocal and style controls, and Nano Banana cover art.
  • Access is rolling out globally in beta for users 18 and older, available on desktop today with mobile arriving over the next few days in English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese.
  • Every track includes an imperceptible SynthID watermark, and Gemini can analyze uploaded audio to determine if it was created with Google’s AI.
  • Google says the feature is designed for original expression, applies filters and reporting tools to reduce direct artist mimicry, and acknowledges the safeguards are not foolproof.
  • Lyria 3 is also being added to YouTube’s Dream Track for Shorts creators, and paid Google AI tiers carry higher generation limits, with some outlets reporting daily caps for free and paid users.