Google rolls out Lyria 3 music generation in the Gemini app

Google has begun rolling out Lyria 3 in the Gemini app, adding a built-in tool for generating short music tracks. The company says the feature is available in beta, and it can produce 30-second tracks from prompts or uploaded media. ^[1][2]

Google says the rollout starts on desktop first, with the mobile app following over the next several days. ^[1]

YouTube video

YouTube

Context and background

Lyria 3 is part of Google DeepMind’s Lyria family of generative music models. DeepMind describes Lyria 3 as its most advanced AI music generation tool and positions it for creating high-fidelity music from prompts. ^[3]

This launch puts Lyria 3 into Gemini’s consumer interface, instead of limiting usage to developer or specialist environments. ^[1][3]

Key details

What launched: Gemini now includes a “Create music” capability powered by Lyria 3, described by Google as its latest generative music model. ^[1][2]

How users generate music: Google says users can generate tracks by describing what they want, such as a genre, mood, or memory. It also supports using uploaded media as creative input. ^[1][4]

Inputs supported:

• Google’s help documentation says users can use text plus uploaded photos or videos as context. ^[4]

• Google’s announcement also describes photo and video uploads being used to inspire a track. ^[1]

Output format: Google says Gemini generates 30-second tracks, and the Workspace update describes them as “custom 30 second soundtracks.” ^[1][2]

Lyrics behavior: Google says users do not need to provide lyrics, because Gemini can generate lyrics from the prompt, while also supporting instrumental outputs. ^[1][4] DeepMind’s prompt guide also explains how to include custom lyrics by adding Lyrics: to a prompt, or how to request lyric generation by making the theme clear. ^[5]

Cover art: Google says Gemini generates custom cover art for each track. ^[1]

Availability and languages: Google says Lyria 3 in Gemini is available for users 18+ in English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese, with plans to expand language coverage. ^[1][4]

Watermarking and provenance: Google says tracks generated in Gemini are embedded with SynthID, and the Gemini app’s verification capabilities are being broadened to include audio. ^[1]

Music-community and copyright posture: Google says music generation with Lyria 3 is designed for original expression rather than mimicking existing artists, and it describes output filtering intended to check against existing content. ^[1]

Related product surface: Google says creators can also explore Lyria 3 through YouTube’s Dream Track, with availability rolling out beyond the U.S. to more countries. ^[1]

I tested the feature on Gemini with the following prompt: “An ambient-classical hybrid piece with a peaceful, contemplative mood, featuring soft felt piano and sustained string ensemble.”

This was the result:

Why this matters

Google is making music generation part of a general-purpose assistant workflow, rather than a separate music tool. That makes it easier for non-specialists to create short soundtracks for personal projects, drafts, and social sharing. ^[1][4]

The launch also highlights a shift in consumer AI products toward multi-modal creation, where text, images, and now audio are generated in the same interface. ^[1]

Brief outlook

Google says it plans to expand language coverage and continues rolling out access across platforms. What is still unclear is whether Gemini will add longer track lengths or deeper editing controls beyond prompt-based generation, since Google has not committed to those capabilities in today’s release materials. ^[1][4]