Lyria 3 is Google DeepMind’s music model, released February 18, 2026 as the successor to Lyria 2 (May 2025). It generates high-fidelity 48kHz stereo audio up to about three minutes, working directly on audio tokens rather than spectrograms, and folds in DeepMind’s acquisition of ProducerAI. SynthID watermarking is mandatory for provenance.
Access is through Vertex AI and the Gemini API, which makes it a natural fit for teams already inside Google’s ecosystem — generating background and production music, or building on Lyria RealTime for live, controllable generation.
Its limitations are reach and emphasis rather than raw quality: lyric and vocal handling are lighter than Suno or Udio, and access is tied to Google’s platforms rather than offered as a standalone consumer studio. Choose it when integration, real-time control and provenance matter more than community and standalone tooling.
Best for
- Background and production music inside Google tools
- Enterprise music generation on Vertex AI
- Real-time/interactive music (Lyria RealTime)
Pros & cons
Strengths
- High-fidelity 48kHz stereo, up to ~3 minutes
- Integrates with Vertex AI and the Gemini API
- SynthID watermarking for provenance; Lyria RealTime for live use
Limitations
- Lyric/vocal handling is lighter than Suno/Udio
- Access tied to Google's platforms