Photo Credit: Alban_Carguy
Google has added its newly released Lyria 3 music generator to Gemini, which can now determine whether audio uploads contain the SynthID watermark.
The tech giant debuted Lyria 3 today, touting the DeepMind-developed model as its most advanced music offering to date. At the top level, this refers to automatic prompt-based lyric generation, bolstered control over outputs, and the ability to pump out “more realistic and musically complex tracks.”
Against the backdrop of a rapidly evolving AI sector – and increasingly intense competition on the music-generation side – these improvements don’t necessarily come as a surprise. But they’re particularly noteworthy given the integration into Gemini, which reportedly boasts in excess of 750 million users.
(Technically, Lyria 3’s beta has already hit Gemini on desktop but is expected to make its way onto mobile over “the next few days for all users globally.”)
Now, said users can utilize text prompts, photos, or videos to produce “30-second tracks with custom cover art” via Lyria 3. And just in case one’s still drawing a prompting blank, navigating to the appropriate tab (“create music,” that is) brings up a variety of pre-made tracks, grouped by genre, that can then be “remixed.”
In other words, this is a big upgrade and an aggressive push for Lyria, which, per Google, is “designed to be your musical collaborator.”
Admittedly, the latter billing might clash with the claim (likewise from Google) that “[t]he goal of these tracks isn’t to create a musical masterpiece, but rather to give you a fun, unique way to express yourself.”
And it also raises questions about the long-term trajectory of Lyria, which is already being framed (“transform your images into high-fidelity audio to express your brand through music”) as well-suited for generating cleared audio to accompany businesses’ social posts.
(What about audio fidelity, anyway? Per Gemini, Lyria 3, trained on “2+ million tracks” versus approximately 500,000 for Lyria 2, upgraded to 24-bit – meaning it’s topped several leading DSPs, including YouTube Music, in the bit-depth department.)
More immediately, though, YouTube’s Lyria-powered Dream Track is expanding beyond the U.S. and into markets around the globe, Google reiterated today.
That tool is tailored for Shorts and, with the majors’ support, has been pushing the artist-soundalike envelope since at least 2023.
Finally, Gemini can now analyze audio (on top of image and video identification support) for the initially mentioned SynthID. One need only upload a file and ask if it was generated by Google AI; Gemini will check for the “imperceptible watermark” and answer accordingly.
