Google adds Gemini AI tool to generate short music tracks

Google has added a music-generation feature to the Gemini app, letting users create 30-second tracks from text prompts or uploaded images in a beta release.

The feature uses Lyria 3, a generative music model from Google DeepMind, expanding Gemini's creative tools beyond image and video generation.

Lyria 3 produces short pieces of music, with optional vocals and lyrics, based on a prompt. Users can also ask Gemini to draw inspiration from an uploaded photo or video and generate a track that matches the mood and subject matter.

How it works

In the Gemini app, users type a description of a genre, mood, memory, or joke. Gemini then generates either a lyrical track or an instrumental. Google positions it as a quick way to create a shareable soundtrack, not a full-length composition.

The app also supports prompt-driven music creation from media uploads. Users can upload photos or a short video and ask Gemini to compose music based on the content. The output includes custom cover art generated by a model called Nano Banana. Tracks can be downloaded or shared via a link.

Google says Lyria 3 improves on earlier versions in three areas: it can generate lyrics without users writing them, offers more control over elements such as style, vocals, and tempo, and produces more realistic, musically complex tracks.

YouTube link

Lyria 3 is also available through YouTube's Dream Track, which generates soundtracks for Shorts. It is available in the US and is rolling out to YouTube creators in other countries.

The YouTube version focuses on short-form audio that can work as a backing track or a lyrical element. Google says the updated model will bring greater customisation for Shorts soundtracks.

Verification tools

Alongside music generation, Google is expanding Gemini's content-verification features to cover audio. The company already uses SynthID, an imperceptible watermarking system, to identify AI-generated content in other formats.

All tracks generated in Gemini will include a SynthID watermark, Google says. Gemini's verification tool also lets users upload an audio file and ask whether it was generated using Google AI. It checks for a SynthID watermark and then uses its own reasoning to respond.

The move comes as creators and platforms face pressure to label AI-generated media and improve provenance signals. Audio has been a particular challenge because it can be edited or re-recorded while still retaining hallmarks of synthetic generation.

Copyright approach

Google says it has developed Lyria in collaboration with the music community since the model's initial launch in 2023. It pointed to its work on Music AI Sandbox and said it considered copyright issues and partner agreements in Lyria 3's training process.

The company says the feature is intended for original expression, not for mimicking existing artists. If a user names a specific artist in a prompt, Gemini treats it as broad creative inspiration and generates a track with a similar style or mood.

Google also says it uses filters that check outputs against existing content, and that users can report content that may violate rights.

Availability

Lyria 3 music generation is available in the Gemini app for users aged 18 and over. It supports English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese.

The rollout starts on desktop, with mobile availability following over the next several days. Google says AI Plus, Pro, and Ultra subscribers will get higher usage limits than free users.

Google frames the feature as a casual creative tool for everyday sharing and social posts, rather than a production suite aimed at professional musicians.

Google says it plans to expand language quality and coverage over time as the feature moves beyond the initial beta.