Google Rolls Out Lyria 3.5 Music Generation to Gemini App and API

Google has integrated its Lyria 3.5 music generation model into the Gemini app and API globally on web and mobile platforms. [1] The model, which previously debuted in Google Flow Music, enables users to generate full songs reaching lengths of two to three minutes. [1][3]

In the Gemini app, the update includes guided creation templates for projects like brand jingles, personalized ringtones, or background music. [1] Users can describe a genre, select vocal styles, or upload an image to prompt track generation. [1]

Through the Gemini API, developers gain deeper compositional control over track creation. [1][3] Users can upload up to ten images to guide the audio generation process. [3]

Additionally, the API allows developers to structure lyrics using specific section tags and apply timestamps to signal exactly when certain instruments should be introduced. [3] The system produces 44.1 kHz stereo audio and incorporates SynthID watermarking into the generated tracks. [3] The model is also available to developers via Google AI Studio and Google Vids. [1]

This rollout signals Google's move to push high-fidelity music generation beyond the Google Flow Music tool where it initially launched. By offering 44.1 kHz stereo audio and specific timeline controls for instruments, Lyria 3.5 provides granular options for developers using Google AI Studio. Moving the model into the Gemini API standardizes music generation alongside text and image capabilities within Google's primary development ecosystem. What the sources don't address: How the SynthID watermarking performs against deliberate tampering or compression when these tracks are distributed externally.