Google Launches Lyria 3 AI Music Generator in Gemini App
Tony Kim Feb 18, 2026 16:27
Google DeepMind rolls out Lyria 3 beta, enabling users to generate 30-second music tracks from text prompts or photos in the Gemini app across 8 languages.
Google DeepMind has released Lyria 3, its latest generative music model, in beta across the Gemini app starting February 18, 2026. The tool lets users create 30-second music tracks by typing a text description or uploading a photo—a significant expansion of Gemini's creative capabilities beyond images and video.
The model generates both instrumental tracks and songs with auto-generated lyrics. Users can specify genre, mood, tempo, and vocal style through natural language prompts. Type something like "a comical R&B slow jam about a sock finding their match" and Lyria 3 produces a complete track within seconds.
What's New in Lyria 3
Three core improvements separate this release from earlier Lyria versions. First, the model now writes lyrics automatically based on user prompts—no need to craft verses yourself. Second, users get finer control over musical elements including style, vocals, and tempo. Third, Google claims the outputs are "more realistic and musically complex" than previous iterations.
The photo-to-music feature stands out as particularly novel. Upload vacation photos and ask for a track about the trip, and Lyria 3 analyzes the visual content to compose something that fits the mood. Each generated track comes with custom cover art created by Google's Nano Banana image model.
Availability and Limits
Lyria 3 is rolling out to users 18 and older in eight languages: English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese. An Arabic beta launched simultaneously. Desktop access started February 18, with mobile app availability following over subsequent days.
Google AI Plus, Pro, and Ultra subscribers get higher generation limits, though Google hasn't specified exact caps for free users.
Copyright Safeguards
Google has built several guardrails into the system. All generated tracks carry SynthID, the company's imperceptible watermark for identifying AI-created content. The Gemini app can now verify whether audio files were generated using Google AI—a capability that previously existed only for images and video.
When users name specific artists in prompts, Lyria 3 treats this as "broad creative inspiration" rather than attempting to replicate that artist's voice or style. Filters check outputs against existing copyrighted material, though Google acknowledges the system "might not be foolproof" and has set up a reporting mechanism for potential violations.
YouTube Integration
Creators on YouTube can access Lyria 3 through Dream Track, the platform's AI soundtrack tool for Shorts. Previously limited to U.S. creators, the feature is now expanding internationally. The integration aims to let Shorts creators generate custom backing tracks or lyrical verses tailored to their content.
Google emphasized that Lyria 3 tracks are meant for personal expression and sharing rather than professional music production. The 30-second limit reinforces this positioning—useful for social content and messaging, but not for replacing studio recordings.
Image source: Shutterstock- ai
- gemini
- lyria 3
- deepmind
