Google launched Lyria 3 Pro on Wednesday, upgrading its AI music generator from 30-second clips to full three-minute tracks. The jump is significant: Lyria 3, which debuted in February, was fun for novelty snippets but useless for anything resembling a real song.
The Pro model now understands song structure. Users can prompt for intros, verses, choruses, and bridges separately, controlling the arc of a composition rather than getting a single undifferentiated block of audio. Google calls this "structural awareness," and it's the feature that moves Lyria from toy to tool, at least in theory. Independent testing hasn't confirmed how well the structural prompting actually works in practice.
Availability spans six platforms: paid Gemini subscribers get it first (10 tracks per day on AI Plus, 20 on Pro, 50 on Ultra), with rollouts to Google Vids, Vertex AI for enterprise clients, and ProducerAI, the formerly-Riffusion startup Google acquired in February. Developers can access it through AI Studio and the Gemini API.
All outputs carry SynthID watermarks. Google says the model was trained on licensed data and won't mimic specific artists, positioning itself carefully while competitors Suno and Udio face ongoing copyright lawsuits from major labels. Pricing for the consumer tiers wasn't separately disclosed; it's bundled into existing Gemini subscription plans.
Bottom Line
Lyria 3 Pro gives paid Gemini subscribers AI-generated music tracks six times longer than the previous 30-second limit, with per-section structural control across six Google platforms.
Quick Facts
- Track length: up to 3 minutes (previously 30 seconds)
- Available on 6 platforms: Gemini app, Google Vids, ProducerAI, Vertex AI, Gemini API, AI Studio
- Daily limits: 10 tracks (AI Plus), 20 (Pro), 50 (Ultra)
- All outputs watermarked with SynthID
- Lyria 3 launched February 2026; Pro follows one month later
