If you’ve ever wanted to make music but have neither the talent nor the inspiration, Google has the AI tool for you. Gemini will now generate a 30-second song for you directly from a text prompt, photo, or video.
Google launched Lyria 3, the latest iteration of its music creation AI, on Wednesday, and has made it far more available than the previous versions of the engine. Like image creation tool Nano Banana and video-making AI Veo, Lyria 3 lives right in the Gemini Tools menu, where it can be selected and used to create a song with little more than a short description.
Google’s examples in its announcement post include an R&B song about socks finding their matches in a washing machine and an afrobeat track about childhood memories cooking plantain-based meals with one’s mother, both of which included lyrics that Google appeared proud to say were written on the fly by AI.
“No need to provide your own lyrics! They’ll be generated for you based on your prompt,” senior product managers Joël Yawili (Gemini) and Myriam Hamed Torres (Google DeepMind) wrote in the announcement before unironically adding that “the goal of these tracks [is] to give you a fun, unique way to express yourself.”
Because nothing says “I love you, Mom” like a song written and performed by a Google AI for the low, low cost of ruining the environment. Oh, and Google is also throwing in cover art generated by Nano Banana, so no need to actually find a sentimental picture to include.
Lyria first launched in 2023, with Google later deploying it in YouTube experiments such as Dream Track, which let creators generate 30-second soundtracks for YouTube Shorts in the style of participating artists, and in Music AI Sandbox, a separate set of tools aimed at musicians looking to sketch and iterate on ideas with AI assistance.
According to Wednesday’s announcement, Lyria 3 improves on previous Lyria models by adding the aforementioned lyric generation feature, as well as giving prompters more control over style, vocals, and other particulars. Lyria 3 can also “create more realistic and musically complex tracks,” Google said. Dream Track, naturally, is being updated with Lyria 3, though it’s not clear whether YouTube creators will be able to spin tracks longer than 30 seconds, which Google said is Lyria 3’s current limit.
Google hopes to avoid a copyright fight
Lyria’s original iteration, as paired with Dream Track, included training data from licensed artists like T-Pain, Demi Lovato, Sia, and other pop middleweights. Google didn’t mention who it may have signed partnership agreements with to train Lyria 3, but did say that it programmed the model to be “very mindful of copyright” and agreements it signed with musicians who offered up their intellectual property as meat for its AI grinder.
“Music generation with Lyria 3 is designed for original expression, not for mimicking existing artists,” Google explained. “If your prompt names a specific artist, Gemini will take this as broad creative inspiration and create a track that shares a similar style or mood.”
There are also filters in place to prevent Lyria 3 from creating something too close to an extant song, Google explained, but it admitted those filters might not get everything right, and asked users to report content that might be a copyright infringement.
For those who wish to experiment with generating a bit of soulless, 30-second musical AI slop, Lyria 3 is rolling out globally beginning February 18 for all Gemini users 18+, and is available now in English, German, Spanish, French, Hindi, Japanese, Korean, and Portuguese with additional languages to be added. It’s launching on desktop first, though not everyone appears to have access yet (it’s not showing up for this vulture), with the Gemini mobile app set to get access over the next several days. ®
First Appeared on
Source link
Leave feedback about this