Google Releases Lyria 3 AI Music Generator

Lyria 3 can generate a 30-second track complete with vocals, lyrics, and music.

Google just dropped a new version of its AI music generator tool, Lyria. This new version is capable of producing a 30-second track from various inputs like text descriptions, photos, and even videos.

The AI music scene has mainly been nominated by platforms like Suno and Udio. Since their explosive rise in 2024, they’ve set the bar high:

• Suno is really good at quick, catchy full-length songs

• The recent version of the music model can produce up to 8+ minutes per track with a strong pop/viral appeal

• Udio often edges out in raw audio fidelity, vocal realism, and studio-like controls (e.g., inpainting, extensions for longer tracks up to 15 minutes).

Lyria 3, in contrast, can only produce a 30-second track per generation. But according to Google, this model creates high-fidelity AI music complete with instrumentals, vocals, and lyrics.

How Lyria 3 compares to the previous versions

The original Lyria model was launched in late 2023 as part of Google DeepMind’s experiments, like Music AI Sandbox (for pro musicians) and YouTube’s Dream Track feature.

Lyria 2 followed around 2025, adding refinements like better control, real-time capabilities (Lyria RealTime), and API access via Vertex AI. It was mostly instrumental-focused and geared toward developers or integrated tools.

And now we have Lyria 3 in 2026.

Here are the main improvements of Lyria 3 over the previous versions: