Gemini adds music generation to its AI app

Google has expanded the creative capabilities of its Gemini app with the introduction of built-in music generation, marking the company’s first consumer-facing step into AI-generated audio. The new feature is powered by Lyria 3, the latest generative music model developed by Google DeepMind, and is rolling out in beta from 18 February.

Music generation builds on Gemini’s existing image and video tools, extending the app’s role as a multi-modal creative assistant. Users can now create original 30-second music tracks by entering a text prompt or uploading an image or video for inspiration. The system translates those inputs into complete audio clips within seconds, producing both instrumental tracks and songs with automatically generated lyrics.

The feature is positioned as a lightweight, expressive tool rather than a professional music production platform. Google frames the capability as a way for users to add personalised soundtracks to moments, jokes, or memories, rather than a path to polished commercial releases. Each track is generated with accompanying cover art created by an internal image model, making the output easy to download or share directly from the app.

How Lyria 3 changes AI music generation

Lyria 3 represents a significant update over Google’s earlier audio models, with changes focused on accessibility and creative control. Unlike earlier tools that required users to supply their own lyrics, the new model can generate lyrics automatically based on the prompt, lowering the barrier for non-musicians.

The model also allows more precise guidance over musical attributes such as style, vocal presence, and tempo. Users can request specific moods or genres, from upbeat afrobeat tracks to slower R&B-inspired songs, and the system adapts its output accordingly. Google says the model produces more realistic and musically complex compositions than previous versions, with improved structure and coherence across the short runtime.

Two primary workflows are supported. Text-to-track generation allows users to describe a theme, memory, or feeling and receive a corresponding piece of music. Image- and video-to-track generation uses visual content as contextual input, with Gemini analysing the uploaded media to infer tone and mood before composing a matching soundtrack. In both cases, the output is capped at 30 seconds, reinforcing the feature’s emphasis on quick creation and sharing.

Creator tools, safeguards, and availability

Beyond the Gemini app, Lyria 3 is also being integrated into creator-facing products. The model is now available within YouTube’s Dream Track for Shorts, initially in the United States and expanding to other markets. The update aims to give creators more flexibility when generating backing tracks or lyrical snippets for short-form video, improving audio quality and customisation without requiring external music libraries.

To address transparency and misuse concerns, all audio generated in the Gemini app is embedded with SynthID, Google’s imperceptible watermarking technology for AI-generated content. Gemini’s verification tools have also been expanded to include audio, allowing users to upload a file and check whether it was created using Google AI systems. This builds on existing image and video verification features within the app.

Google says Lyria 3 has been developed with an emphasis on responsible use and respect for the music ecosystem. The system is designed to generate original compositions rather than imitate specific artists. Prompts that name real musicians are treated as broad stylistic inspiration, and outputs are checked against existing content to reduce the risk of duplication. Users can report material they believe infringes on rights, and use of the feature is governed by Google’s general AI usage policies.

Lyria 3 is available in the Gemini app for users aged 18 and above, with initial language support covering English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese. The feature is rolling out on desktop first, followed by mobile over the coming days. Higher usage limits are reserved for paid Gemini subscription tiers.