Google DeepMind shipped Lyria 3, its latest generative music model, inside the Gemini app today. Users can create 30-second tracks by describing an idea or uploading a photo.

The model generates lyrics automatically based on prompts, eliminating the need for users to write their own. Lyria 3 also offers enhanced creative control over style, vocals, and tempo compared to previous versions.

Users can create tracks through two methods: text-to-track generation using genre, mood, or memory descriptions, or photo-to-track creation where uploaded images inspire musical compositions. The app generates custom cover art alongside each track.

Enhanced capabilities and watermarking

Lyria 3 represents a significant upgrade from Google's earlier Lyria models. The system can produce more musically complex tracks with realistic audio generation capabilities.

All generated music includes SynthID, Google's imperceptible watermark for AI-generated content. The Gemini app now supports audio verification alongside existing image and video detection tools.

Users can upload audio files to check if they were created using Google AI systems. The verification process combines SynthID detection with Gemini's reasoning capabilities.

Broader rollout across platforms

The feature launches in beta for users aged 18 and above across multiple languages. Gemini subscribers receive higher generation limits than free-tier users.

Lyria 3 also powers YouTube's Dream Track feature for Shorts creators. The model is rolling out to YouTube creators in the US and expanding to other countries, enhancing soundtrack customization for short-form video content.

Google positions the tool as a creative expression medium rather than a professional music production platform. The 30-second format targets quick sharing and personal creativity over commercial music creation.