Google is adding a music-generation feature to the Gemini app, expanding its AI toolkit beyond text and images into short-form audio creation. The feature, announced Wednesday, Feb. 18, 2026, is powered by DeepMind’s Lyria 3 model and is launching in beta.
The workflow is simple: describe the kind of song you want, and Gemini generates a 30-second track—with lyrics by default (or instrumental audio if you ask for it). Google’s own example is delightfully specific: request a “comical R&B slow jam about a sock finding its match,” and Gemini will produce a short clip plus custom cover art generated by Nano Banana.
👉 Browse music production & songwriting books on Amazon
(Affiliate link: using this link may earn a commission.)
Turn photos and videos into songs
Google says Lyria 3 in Gemini can also take a photo or video upload and generate music that matches the mood of the media. That’s a notable shift from “text-to-music” toward something closer to an automated soundtrack tool—useful for creators who want quick background music for short videos, slideshows, or reels.
Lyria 3 is pitched as an upgrade over earlier models, producing more realistic and complex tracks while giving users additional controls over elements like style, vocals, and tempo.
YouTube creators get Lyria 3 via Dream Track
Alongside Gemini, Google is also pushing Lyria 3 to YouTube creators through Dream Track, YouTube’s AI music feature. Google says Dream Track was previously available in the U.S., and it’s now rolling out to creators in other countries—an expansion, though not explicitly framed as a complete “global” launch.
Artist style prompts are allowed—within guardrails
Google is taking a familiar stance on “style.” The company says music generation is meant for original expression, not direct imitation. If a prompt names a specific artist, Gemini will treat it as broad creative inspiration and generate something with a similar vibe—not a clone—while applying filters that check outputs against existing content.
To help with transparency, Google says every track generated with Lyria 3 will include a SynthID watermark to identify it as AI-generated. Gemini will also gain the ability to check uploaded audio for SynthID and answer whether it’s AI-generated (with additional reasoning as needed).
Availability and languages
The feature is rolling out to Gemini users 18+ with support for English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese. Google notes that it’s rolling out on desktop first, with mobile support following over the next several days.
The bigger picture: AI music is still a live wire
Music generation remains one of the most contentious corners of generative AI. Platforms and tech companies are experimenting with ways to integrate AI audio, while the music industry continues to push back—particularly through lawsuits over copyrighted training data and demands for licensing frameworks. Meanwhile, streaming services and platforms are building detection and labeling tools to curb abuse, including fraudulent uploads and artificial streaming behavior.
Google’s bet with Lyria 3 is clear: make creation frictionless, add transparency signals like SynthID, and expand access—while hoping the guardrails hold up in the real world.