Google Brings AI Music Technology to Gemini—We Tried It, and It's Too Little Too Late – Decrypt

Google has spent years quietly constructing its AI music mannequin. On Tuesday, it lastly put it someplace everybody might truly use.

Lyria 3, Google DeepMind’s newest music era mannequin, is now rolling out in beta contained in the Gemini app, letting any person over 18 describe an thought or add a photograph and get a completely produced monitor again in seconds—lyrics, instrumentals, and AI-generated cowl artwork included.

“Simply describe an thought or add a photograph, like ‘a comical R&B sluggish jam a couple of sock discovering their match’ and in a matter of seconds, Gemini will translate it right into a high-quality, catchy monitor,” Google mentioned in its official weblog. “To push the artistic envelope additional, you may even ask Gemini to take inspiration from one thing you add.”

We gave it a spin. The quick model: It really works, it is enjoyable, and it might impress anybody who has by no means used different cutting-edge fashions like Suno or Udio. For many who have, it isn’t going to interchange their workflows anytime quickly.

The tracks Lyria 3 produces are 30 seconds lengthy. That is the ceiling proper now, and Google is upfront about it—the acknowledged aim is not to make polished industrial songs however to create shareable moments. The output we examined was coherent, the lyrics matched the immediate, and the manufacturing high quality was first rate sufficient.

Here’s what Google, Suno, and Udio can do with the identical immediate:

The difficulty begins if you push the perimeters. Lyria 3 appears to have a snug vary of genres it navigates nicely—pop, afrobeat, R&B, mild hip-hop. However once we examined prompts calling for extra particular or uncommon types, the mannequin had a tough time sticking to them. The prompting information from DeepMind acknowledges this implicitly: It is heavy on examples for mainstream genres and lightweight on steerage for something exterior these lanes.

Examine that to Udio, which when it launched in 2024 got here with controls like a “immediate energy” slider, a readability adjuster for background noise, and unfavorable prompting to exclude particular sounds or types. Or Suno, which has been producing full multi-minute songs with correct track construction—verses, choruses, bridges—and is taken into account one of the best mannequin within the scene. Each instruments allow you to generate long-form tracks that really feel like precise songs, not samples. Lyria 3’s 30-second cap and tendency to float from uncommon prompts put it in a distinct tier.

There’s additionally some authorized context value mentioning. Each Suno and Udio have been sued by the Recording Business Affiliation of America in 2024 for allegedly coaching their fashions on copyrighted recordings with out permission. Udio settled with Warner Music in November 2025 and is now transitioning to a completely licensed platform, launching someday in 2026. Suno’s case continues.

Google, for its half, says it has been “very aware of copyright and accomplice agreements” in coaching Lyria 3, and the mannequin intentionally avoids mimicking particular artists—should you identify one in a immediate, Lyria takes it as temper inspiration fairly than a direct instruction.

On the security and transparency aspect, all tracks generated in Gemini come embedded with SynthID, Google’s imperceptible AI watermark. The corporate additionally added audio verification to Gemini: you may add a monitor and ask whether or not it was made by Google AI. That form of provenance tooling is more and more vital as AI-generated audio floods streaming platforms—Deezer has already deployed detection instruments to determine and flag faux streams from AI music.

Lyria 3 is offered as we speak on desktop for all Gemini customers 18 and older in English, and a handful of extra languages. Cellular is rolling out over the following few days. Google AI Plus, Professional, and Extremely subscribers get increased era limits. The mannequin can also be increasing YouTube’s Dream Monitor function globally, beforehand U.S.-only, giving Shorts creators entry to AI-generated soundtracks for his or her movies.