Google’s Lyria 3 Arrives in Gemini for Custom Music Creation

Google is putting a music studio inside Gemini. The app now lets users generate 30-second AI tracks on demand, turning simple prompts into fully formed songs.

The update brings DeepMind’s Lyria 3 model directly into Gemini, allowing text or photos to be transformed into custom music with generated lyrics and shareable audio, according to a Google announcement.

Text in, music out

Users can describe a genre, mood, memory, or inside joke, and the system composes an original track complete with vocals or instrumental backing. There’s no need to supply lyrics; they can be generated automatically based on the prompt.

Uploads add another layer. A photo or short video can be used as creative input, with the model building a soundtrack that matches the tone of the scene. Prompts can also specify elements like tempo, vocal style, and overall feel, giving users more control over how the final track comes together.

Built to share in seconds

Each generated track arrives with custom cover art, giving users a ready-made visual to match the audio. Songs can be downloaded directly or shared through a link, keeping the process lightweight and immediate.

The rollout begins on desktop, with mobile availability following over the next several days. Access is open to users 18 and older in supported languages, while Google AI Plus, Pro, and Ultra subscribers receive higher generation limits.

The feature is available in the US and is expanding to creators in additional countries, extending music generation beyond the Gemini app and into YouTube’s broader creation tools.

Built-in detection for AI-generated audio

All tracks created in the app are embedded with SynthID, Google’s imperceptible watermarking technology designed to identify AI-generated content. The marker is woven directly into the audio at creation, allowing it to be detected later without affecting the music’s sound.

Gemini now also supports audio verification. Users can upload a file and ask whether it was generated 

From Gemini to Shorts

Lyria 3 is also powering YouTube’s Dream Track feature, giving creators new ways to generate custom soundtracks for Shorts. The upgrade improves audio quality and allows for more tailored lyrical verses or instrumental backing to match short-form video. Using Google AI, the system will check for SynthID while applying its own analysis before returning a result.

Style without impersonation

Music generation is built for original work, not replication. When a prompt names a specific artist, the system draws on broad stylistic cues instead of copying a voice or catalog.

Google said Lyria 3 was developed with attention to copyright and partner agreements, with filtering for checking outputs against existing works. Users can report potential rights violations, and the feature is governed by the company’s Terms of Service and generative AI policies.

Not every institution is embracing built-in AI, with the European Parliament moving to restrict it on staff hardware.