Lyria 3: Vocals, Lyrics, and the Next Generation

Gemini app (all tiers)2 min read

Lyria model evolution

Lyria 1

Nov 2023

Initial model. Instrumental music generation. Powered early MusicFX experiments.

Lyria 2

2024

Improved quality. Available on Vertex AI API. 32.8 second instrumental clips at $0.06/30s.

Lyria RealTime

Mid 2025

Real time streaming model. Powers MusicFX DJ. Continuous, steerable music at 48kHz.

Lyria 3

Feb 2026

Vocals and lyrics generation. Image/video to soundtrack. Available in Gemini app.

Lyria 3 capabilities

🎤
Vocals and Lyrics

Generate complete songs with human sounding vocals and coherent lyrics matched to your requested genre and theme.

🖼️
Image to Soundtrack

Upload a photo and Lyria 3 analyzes the visual mood, content, and style to generate a matching musical accompaniment.

🎬
Video to Soundtrack

Share a video clip and receive a soundtrack that complements the visual action, pacing, and emotional tone.

🌍
Multi Language Vocals

Generate vocals in multiple languages, expanding creative possibilities beyond English only music generation.

🎵
Genre Versatility

Pop, rock, jazz, classical, electronic, hip hop, lo fi, ambient, drum and bass, funk, and many more genres and hybrid styles.

🔒
SynthID Embedded

Every generated track carries an imperceptible SynthID watermark for responsible AI identification and provenance.

The vocal generation breakthrough

Before Lyria 3, AI music generation was almost entirely limited to instrumental tracks. Generating convincing vocals with coherent lyrics that match the musical style was one of the hardest unsolved problems. Lyria 3's ability to produce songs with vocals places it alongside Suno and Udio as one of the few systems that can generate complete songs, not just backing tracks.