On September 4, 2026, Google announced that its Lyria 3.5 music model was coming to the Gemini app and Gemini API. It is no longer limited to a dedicated creative tool: users, developers and video makers can use text or images as the starting point for a track.
What changed in Lyria 3.5?
Google describes Lyria 3.5 as its best-sounding music model, with more expressive vocals, richer arrangements and higher fidelity. In Gemini, users can choose or describe a genre, switch between vocals and instrumental tracks, and start from templates such as background music or birthday songs.
The API documentation is more specific: the model generates 44.1 kHz stereo audio from text or images and supports lyrics plus structures such as verses, choruses and bridges. Developers can provide up to ten images, making it useful for video scoring, brand-music drafts, game prototypes and interactive content.
Where is it available?
Google says Lyria 3.5 is available globally in the Gemini web and mobile apps, as well as Google AI Studio, the Gemini API, Google Vids and Flow Music. The model ID for developers is `lyria-3.5`, and it remains a preview model.
The app and API are not identical. Gemini emphasizes low-friction choices for genre, vocals and length, while the API is a route for embedding generation in a product. Teams should still check account access, regional availability, output formats and current model documentation.
It is not a full music editor yet
The public documentation presents the API as a one-shot generation workflow. If a chorus or lyric direction is wrong, users generally regenerate rather than replace one section as they would in a DAW. Preview limits, pricing and lifecycle may change, so production systems should not assume a permanently stable interface.
Google also keeps safety boundaries that block requests to imitate specific artists' voices or generate copyrighted lyrics. Generated audio carries SynthID. These measures can reduce impersonation and provenance confusion, but they do not settle commercial licensing or responsibility for a release.
What it means for creators and developers
The important change is not that AI can make music, but that full-song generation is now in a general chat interface and a programmable API. Creators can prototype a score faster and developers can connect images, text and audio in a product workflow. Human selection, editing, rights checks and versioning remain necessary.
Google announcement: Create your best tracks yet with Lyria 3.5 in Gemini
