Lyria · 3 / 3.5

Generate music and vocals through Google's hosted creative products and supported APIs.

GenerationCommercial hosted
Start here
Specify instrumentation, vocal character and duration. Check arrangement and lyrics across several takes.
Training data
The Lyria 3 card describes audio annotated with captions at multiple detail levels, with deduplication, quality and safety filtering. No named corpus, hours or track inventory is supplied.
Training rights
The card does not provide recording-level permission evidence. Artist collaborations do not identify the complete training set.
License & output terms
Proprietary hosted access under the selected Google product's terms.
Limitations
Outputs carry SynthID watermarking. The family page lists 3.5; training detail here comes from the 3 card, which covers subsequent versions without a separate corpus breakdown.
Technique
Latent diffusion over temporal audio latents (Lyria 3 card)
Runtime
Hosted products and APIs; distinct from the downloadable Magenta RealTime models.
Parameters
Not itemized for this release/family. Parameter sizes shown elsewhere may exclude encoders, decoders or other runtime components.

Sources reviewed . Training and rights notes summarize provider disclosures. A model license does not clear every source recording or generated output. Parameter counts refer to the named variant or component, not total runtime memory.