Overview
- Lyria 3 Pro, which Google unveiled Wednesday, generates songs up to three minutes with section-level prompts for intros, verses, choruses and bridges.
- Access is rolling out to paid Gemini app tiers with daily caps of 10, 20 or 50 tracks, to developers in public preview via the Gemini API and Google AI Studio, and to enterprises on Vertex AI.
- The model adds granular controls like tempo conditioning and time‑aligned lyrics, supports image‑to‑music prompts, and produces realistic vocals across multiple languages and genres.
- Google says it trained the system on partner content and data it has rights to use, checks outputs against existing works, blocks direct artist imitation, and embeds every track with SynthID watermarks.
- Coverage notes growing risks from impersonation and mass AI uploads on streaming services, with reports citing roughly 50,000 AI tracks posted to Spotify daily and 75 million removed last year as platforms add new detection and review tools.