Google's Lyria 3 Brings AI Music Generation to the Gemini App

Google DeepMind's Lyria 3, its latest generative music model, is rolling out today in beta inside the Gemini app, letting users turn a text prompt, photo or video into a finished 30-second track with lyrics in seconds.

Google frames the feature as a creative-expression tool rather than a production instrument for musicians. "The goal of these tracks isn't to create a musical masterpiece, but rather to give you a fun, unique way to express yourself," the company said in its announcement.

The launch matters beyond novelty. Generative music has become a contested frontier between AI developers, record labels and artists over training data, artist likeness and copyright. Google is using Lyria 3's rollout to signal where it stands: original expression only, no artist mimicry, mandatory watermarking, and training that the company says respects copyright and partner agreements.

What Lyria 3 does

Users describe an idea in words or upload an image, and Gemini composes a track to match. Google's examples include "a comical R&B slow jam about a sock finding their match" and a prompt asking for "a fun afrobeat track with a true African vibe" about childhood memories of a mother's home-cooked plantains. Users can also point Gemini at an upload — "Use these photos to create a track about my dog Duncan on a hike in the woods" — and have the model compose lyrics that fit the mood of the content.

According to Google, Lyria 3 improves on earlier Lyria models in three ways:

• Generated lyrics. Users no longer need to supply their own; the model writes them from the prompt.
• More creative control. Users can steer style, vocals and tempo.
• Higher musical complexity. The model produces what Google calls more realistic and musically complex tracks.

The app pairs every track with custom cover art generated by Nano Banana, Google's image model, and supports sharing via download or link.

Availability

Lyria 3 in the Gemini app is open to all users 18 and older in English, German, Spanish, French, Hindi, Japanese, Korean and Portuguese. Desktop access starts today; the mobile app follows "over the next several days." Google AI Plus, Pro and Ultra subscribers get higher usage limits. Google says it plans to expand both quality and language coverage over time.

For video creators, Lyria 3 is also available on YouTube's Dream Track. The feature is live in the U.S. and now rolling out to YouTube creators in other countries, where Google says it will improve the quality of custom Shorts soundtracks — whether a lyrical verse or a backing track.

Audio verification

Every track generated in the Gemini app carries SynthID, Google's imperceptible watermark for identifying its AI-generated content. Google is also extending the Gemini app's verification capabilities to audio, alongside existing image and video checks. Users can upload a file and ask whether it was generated with Google AI; Gemini will check for SynthID and, using its own reasoning, return a response.

The verification expansion arrives as regulators and platforms press for provenance standards for synthetic media. Audio has lagged images and video in detection tooling, and Google is positioning SynthID checks as a partial answer.

The copyright posture

Google says it has developed the Lyria line "responsibly in collaboration with the music community" since first launching Lyria in 2023, citing experiments such as the Music AI Sandbox and attention to copyright and partner agreements in Lyria 3's training.

The policy details are specific. Music generation with Lyria 3 is "designed for original expression, not for mimicking existing artists." If a prompt names a specific artist, Google says Gemini treats it as broad creative inspiration and produces a track in a similar style or mood — not a replication of the artist's voice or sound. Filters check outputs against existing content. Google acknowledges the approach "might not be foolproof," so users can report content that may violate their rights or the rights of others. Users must also adhere to Google's Terms of Service and Gen AI prohibited use policies, which bar violations of intellectual property and privacy rights.

That stance contrasts with ongoing litigation and disputes elsewhere in the industry, where AI music tools have faced accusations of enabling voice cloning and unauthorized artist imitation. Google's line — style transfer yes, artist mimicry no — is now written directly into the product's behavior.

Why it matters

Lyria 3 puts full music generation, lyrics included, in a mainstream consumer app at no added cost for adult users, weeks after Google made image generation a core Gemini feature. It also moves SynthID from images and video into audio verification, giving Google an end-to-end provenance story across modalities.

The constraints Google has chosen — no artist names honored as prompts, output filters, mandatory watermarking — will define how far casual users can push the tool, and they will be tested at scale as the feature reaches the Gemini app's global user base starting today at gemini.google.com.