Lyria
DeepMind's flagship music model: vocals, lyrics and instrumentals with artist-grade control.
About
Lyria is Google DeepMind's music generation model, built to handle the hard parts of music AI: vocals, lyrics and backing tracks that stay coherent together, with fine control over style and performance. It debuted in November 2023 alongside YouTube's Dream Track experiment, which let Shorts creators generate 30-second songs in the voices of nine participating artists, and a Music AI Incubator exploring tools like humming a melody and getting back a full arrangement.
Lyria 2 followed with an upgraded Music AI Sandbox for musicians and producers plus a real-time variant, and by 2026 the model had reached the Gemini app and Google's AI APIs, putting full-length song generation in front of hundreds of millions of users. Everything Lyria generates carries a SynthID watermark so AI-made audio stays identifiable. The model itself is proprietary with no open weights; access runs through Google's AI and cloud APIs.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
No public benchmarks; evaluated through artist partnerships and user studies (provider-reported).
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: DeepMind flagship with multi-year rollout across YouTube, Music AI Sandbox and the Gemini app; SynthID watermarking on all outputs. Strengths: enormous real-world distribution, artist partnerships, responsible-deployment posture. Docked for: (1) fully closed, no open weights; (2) no public benchmarks, quality judged via provider-run studies. License: proprietary (Google DeepMind).
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 9.0/10
- Maintenance · recency, releases, issue load
- 9.0/10
- Openness · license + docs
- 1.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · Announced Nov 2023 with YouTube Dream Track (30s artist-voiced songs)
- · Lyria 2 plus Music AI Sandbox upgrade; real-time variant
- · In Gemini app and Google AI APIs by 2026 (press-reported)
- · SynthID watermarking on all generated audio
- · No open weights; API-only access; Google DeepMind logo staged