Music Generation
In February 2026, Google added music generation directly into the Gemini app, powered by Google DeepMind's Lyria 3 model. You can describe a mood, genre, or scene — or upload a photo or video — and Gemini generates an original music track for it. All output is watermarked using Google's SynthID technology.
What You Can Generate
Gemini's music feature supports two primary output types:
- Instrumentals: Background music tracks without vocals — useful for videos, presentations, or creative projects.
- Songs with lyrics: Tracks with AI-generated vocals and words matching your prompt. You can specify mood, style, tempo, and instrumentation.
Generated tracks are 30 seconds long. You can generate multiple variations from the same prompt and iterate by refining your description.
How to Use It
Music generation is available directly inside the Gemini app. There is no separate tool to open — just describe what you want in a Gemini chat message. You can prompt in text, or attach an image or short video and ask Gemini to create music that matches the scene or mood.
Prompts are accepted in English, French, German, Spanish, Hindi, Japanese, Korean, and Portuguese. The feature is available globally as of February 2026 across Free and Pro plan tiers.
Lyria 3 Model
Lyria 3 is Google DeepMind's latest music generation model. It grew out of the earlier Music AI Sandbox — a suite of experimental tools that Google ran as a collaboration platform for professional musicians and producers. The Sandbox served as a research and testing ground; Lyria 3 incorporates the learnings from that programme into a model suitable for consumer use.
For developers, music generation via Lyria RealTime is also available through the Gemini API, enabling integration into third-party applications and creative tools.
ProducerAI
In February 2026, Google acquired ProducerAI — a standalone AI music generation platform — and moved it into Google Labs. ProducerAI runs on a preview version of Lyria 3 and is aimed at music producers and creatives who want more professional-level control over generation parameters. It operates as a separate product from the Gemini app's music feature.
SynthID Watermarking
All music generated by Lyria 3 in Gemini is watermarked using SynthID — Google's imperceptible AI-content watermarking technology. The watermark is embedded in the audio signal in a way that persists through common audio processing (compression, re-encoding, pitch shifting) but is inaudible to the human ear. SynthID watermarks can be detected by Google's tools to verify whether audio was AI-generated.
Practical Use Cases
- Background music for YouTube videos, reels, or presentations without licensing concerns
- Lo-fi study music or ambient soundscapes generated on demand
- Quickly prototyping a musical direction for a project before commissioning professional recording
- Generating mood-matched music from travel photos or product videos
Checklist
- When was music generation added to the Gemini app, and which model powers it?
- What are the two types of music output Gemini can generate?
- In which languages can you write music generation prompts?
- What is the relationship between Music AI Sandbox and Lyria 3?
- What is SynthID, and why is it applied to all Gemini-generated music?