This Artificial Intelligence Music From Google Will Drive You Crazy
If you’ve ever played with ChatGPT , you know how impressive (and terrifying) AI can be. ChatGPT can generate just about anything you ask it to, with prompts as simple or complex as you can imagine. Well, Google has figured out how to do the same. But instead of spitting out poetry or computer programs, Google’s AI is writing music, and you can preview its creations right now.
Called MusicLM, Google’s artificial intelligence is designed to play high-quality music for minutes based on text prompts. While AI music isn’t necessarily new, the company claims its program creates tracks that match the original text cue more closely and are of better quality than previous models. Big claims, of course. But judging by these previews , they can match them.
The first batch demonstrates how MusicLM can generate music from extended credits, similar to how OpenAI’s ChatGPT generates its famous user responses. Google can pass a title to MusicLM, such as “Main soundtrack for an arcade game. It’s dynamic and upbeat, with a catchy electric guitar riff. The music is repetitive and easy to remember, but with unexpected sounds like crashing cymbals or drumming,” and the AI will process a track that meets those criteria.
My personal favorite in this section is a song created from “Funky piece” with a strong, danceable beat and a prominent bass line. A catchy keyboard melody adds richness and complexity to the song” on the second page, which sounds like something out of Stardew Valley.
However, the uncanny valley actually kicks in whenever Google asks MusicLM to produce vocals. Many of the Google examples here have AI vocals, and while some of them sound “computerized” for lack of a better word, others sound too close to home. When you listen with the full knowledge that the voice is 100% artificial, it’s a little unsettling. Google even asked MusicLM to generate a rap track, and to be honest, “rap” is eerily realistic, even if none of the words they read are real.
If you really want to be scared, scroll down to “Word Processing & Melody”. Here you can hear a series of voices and instruments doing their best with different melodies and the results are just wild. In the very first version, MusicLM sings “Bella Ciao” and it sounds too good. Note: The Tribal Drums and Flute version gives White Lotus.
We also have Long Generation, which creates five-minute long tracks from short descriptions such as “melodic techno” or “relaxing jazz”. But things get really interesting in Story Mode. Google passes several signatures to MusicML and the AI tunes the song based on the given signature. It starts with 15 seconds based on “time to meditate” before the title switches to “time to wake up”. The music fades into the melody as if it’s starting a new verse, with some creepy vocals to boot.
Besides the creepy vocals, it’s interesting to watch the AI change the song without changing the song itself. For the most part, it incorporates the new title into the song without feeling like you’re jumping into a new tune at all. Particularly impressive is the third example, in which the AI jumps from a “pop song” to a “rock song” to a “death metal song”. While the “rock song” doesn’t really sound like rock, the “death metal song” is just fantastic. I believe that AI Death will be a big hit among metalheads.
Another interesting experiment Google did was to give MusicLM descriptions of famous paintings to see what kind of music it would create. If you’ve ever wanted to know what Vincent van Gogh’s “Starry Night” would sound like as a song, then this is your chance. It’s also interesting to see how MusicLM generates the sounds of specific instruments and genres. The electric guitar is on point, as is British Indie Rock, complete with a dramatic twist at the end of the preview.
There’s more to explore in this limited preview of MusicLM, from “Accordion Solos” to comparing different results from the same clue. Google has also published a 15-page research paper on MusicLM if you’re interested in the technical details of the system. While I hope the company releases a beta to the general public, looking at these samples is the next best thing.
[ Fringe ]