freycreate
By Google DeepMind·March 25, 2026

Lyria 3 Pro

Google DeepMind (London / Mountain View, USA) — Google's premium AI music generator. Original music up to 3 minutes with vocals, defined verse/chorus structure, and 44.1 kHz stereo quality.

Lengthup to 3 min
Audio44.1 kHz stereo
VocalsMulti-language
StructureVerse + chorus
InputText + image
ReleaseMar 2026

Coffee Jazz – 30s

Strengths

Up to 3 minutes of music

Full tracks with intro, verses, chorus, bridge, and outro — not just 30-second loops. The Pro variant delivers commercially usable complete songs.

Vocals + lyrical structure

Generates vocals in multiple languages (German, English, Spanish, French, Japanese, and more). Lyria 3 Pro understands musical structure — you can instruct intro, verse, chorus, and bridge explicitly.

Image-to-music

Upload an image and Lyria generates a track that matches its mood, color palette, and atmosphere. A unique workflow for visual-first creators.

44.1 kHz stereo

High-quality audio straight from the model. Structural coherence with vivid instrumental arrangements. SynthID watermark included.

Showcase

Real generations

Coffee Jazz – 30s

Sanfter Jazz als Hintergrundmusik für eine Kaffee-Werbung, warm und einladend

0:000:00

Epic Cinematic – 30s

Epic cinematic orchestral score, sweeping strings, powerful brass, dramatic build-up

0:000:00

Pop Hook – 30s

Energetic pop song with a catchy hook "Today is the day", uplifting major key

0:000:00

Birthday Song (3 min)

Geburstagslied, fröhlich, euphorisch, Reim zeilenweise auf deutsch, mit sich wiederholendem Refrain

0:002:56

Best for

Ad spots and brand sound (background music)

YouTube intros, outros, and social media reels

Complete songs with vocals and structure

Image-to-music for mood-driven composition

Personalized songs (birthdays, events, branded jingles)

Not its strong suit

Max 3 minutes — chain multiple generations for full album tracks

Generation time is longer than image models (typically 30-60 s)

Technical details

Architecture

Lyria 3 is Google DeepMind's music generator (offices in London + Mountain View). Diffusion-based audio architecture with unified vocal/instrumental heads. Lyria 3 Pro understands musical structure (intro, verse, chorus, bridge, outro) and produces structurally coherent long-form tracks up to 3 minutes.

Key features

  • Track length: 30 s (Lyria 3) to 3 min (Lyria 3 Pro)
  • Audio quality: 44.1 kHz stereo
  • Vocals in 8+ languages (DE, EN, ES, FR, HI, JA, KO, PT)
  • Structured prompts: intro / verse / chorus / bridge
  • Image-to-music input
  • SynthID watermark included
  • Use from Claude Code, Cursor & Codex via freycreate MCP
Use directly from AI agents (MCP)

Ready to get started with Lyria 3 Pro?

Beta access is free for selected creators. Claim your slot.

More models

Also interesting