Some or all of the information on this page might not apply to Cloud de Confiance by S3NS. See Differences from Google Cloud for more details.

Lyria 3

Lyria is a music generation model from Google. This page documents the capabilities and features of Lyria 3.

3 Pro Preview

Model ID	`lyria-3-pro-preview`
Modalities	Text Input only Image Input only Audio Output only Video Not supported
Capabilities	Text to music Supported Image to music Supported Vocal generation Supported Instrumental mode Supported Lyrics generation Supported User-provided lyrics Supported Negative prompting Not supported Song generation Full song Supported Detailed structure controls Duration, BPM, Intensity Supported Audio watermarking Supported Filtering Input, output (recitation), output (vocal likeness) Supported Prompt rewriter Supported Content Credentials (C2PA) Supported
Technical specifications	Audio	Maximum audio clip length: 184 seconds Maximum number of clips per prompt: 1 Supported sample rates: 44.1 kHz Supported bitrate: 192 kbps Supported languages: English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese Supported MIME types: `audio/mp3`
Supported regions	Model availability	Global: `global`
Quotas	Regional online prediction requests per minute per base model: 10 tokens per minute
Versions	`lyria-3-pro-preview` Launch stage: Preview Release date: 2026-03-25

Model ID	`lyria-3-clip-preview`
Modalities	Text Input only Image Input only Audio Output only Video Not supported
Capabilities	Text to music Supported Image to music Supported Vocal generation Supported Instrumental mode Supported Lyrics generation Supported User-provided lyrics Supported Negative prompting Not supported Song generation 30 second clips Supported Detailed structure controls BPM, Intensity Supported Audio watermarking Supported Filtering Input, output (recitation), output (vocal likeness) Supported Prompt rewriter Supported Content Credentials (C2PA) Supported
Technical specifications	Audio	Maximum audio clip length: 30 seconds Maximum number of clips per prompt: 1 Supported sample rates: 44.1 kHz Supported bitrate: 192 kbps Supported languages: English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese Supported MIME types: `audio/mp3`
Supported regions	Model availability	Global: `global`
Quotas	Regional online prediction requests per minute per base model: 10 tokens per minute
Versions	`lyria-3-clip-preview` Launch stage: Preview Release date: 2026-03-25

For Lyria pricing information, see the Lyria section of the Cost of building and deploying AI models in Vertex AI page.