Get API access
Model release

Google Releases Lyria 3 Pro Preview for Music Generation

Google announced Lyria 3 Pro Preview on March 25, 2026. This model generates full-length songs from text and image inputs at 48kHz stereo quality.

Released Published 3 min readModel evidence ↗
Lyria 3 Pro Preview release illustration — Google

Official Announcement Details

Google officially launched the Lyria 3 music generation models on March 25, 2026. The announcement identified two specific variants: lyria-3-clip-preview and lyria-3-pro-preview. The source event date is distinct from this article's publication date of August 9, 2026.

  • lyria-3-clip-preview generates 30-second clips.
  • lyria-3-pro-preview generates full-length songs.
  • Both models accept text and image inputs.
  • Output audio is high-quality 48kHz stereo.

Model Specifications and Limitations

The Lyria 3 Pro Preview belongs to the Gemini family and is provided by Google. Its input modalities are listed as text, while its output modality is audio. The model has a context window of 1,048,576 tokens and a maximum output of 65,536 tokens. Open weights are not confirmed for this model. Pricing information is currently unavailable in the source data.

  • Provider: Google.
  • Family: Gemini.
  • Context Window: 1,048,576 tokens.
  • Max Output Tokens: 65,536.
  • Open Weights: Not confirmed.

Practical Evaluation Procedure

Engineers can evaluate the model by running controlled tests using the documented input types. First, test the model with a standard text prompt to generate a full-length song. Measure the resulting audio file to validate that it meets the 48kHz stereo specification. Next, compare the output duration against the expected full-length format versus the 30-second limit of the clip variant. Finally, validate the acceptance of image inputs by providing an image alongside a text prompt and measuring the consistency of the generated audio. These steps confirm functional behavior without inferring performance rankings or architectural details.

Primary sources

QUICK ANSWERS

Frequently asked questions

What inputs does Lyria 3 Pro Preview accept?

According to the official changelog, both Lyria 3 preview models accept text and image inputs. The model description specifically lists text as an input modality, but the launch summary confirms image support for both variants.

Is open-weight access available for this model?

The source data records that open weights are not confirmed. This status does not establish whether local installation is possible or impossible, nor does it define any specific hosting arrangement or API requirement.