Bruno Mars "That's What I Like"

Lots of people dream of being a dancer, but for one reason or another, it never comes to fruition. Perhaps they lack the resources to take classes, or their body just isn't built for the art form. Well, researchers from UC Berkeley have been working on making the dream of being a dancer reality for everyone - not on stage, but on video using Artificial Intelligence technology.

In a new paper titled "Everybody Dance Now" (anybody else going to have that song stuck in their head now?), four researchers detail how they have created an AI program that can take movement from a dancer (the 'source subject') and put it on the body of a non-dancer (the 'target subject'). First, they film a dancer doing some sort of dance style, ranging from ballet to hip-hop. Then, they their actions and create what is essentially a moving stick figure. Using two algorithms called Generative Adversarial Networks, or GAN, the computer program is able to generate a video showing their target subject dancing like they're a seasoned professional!

It doesn't stop there, though. The researchers even figured out how to take movement from music videos (like Bruno Mars's "That's What I Like") and superimpose it on top of their target subjects, giving them the same sweet moves that Bruno has.

It's certainly not perfect - if you watch the video, there are some clear technical glitches and abnormalities. BUT it's still pretty awesome, fairly convincing, and simply fun!

About the author

Anne Luben has performed works by notable choreographers such as Donald McKayle, Bill T. Jones, Jiri Kylian, Idan Cohen, Alex Ketley, and Summer Lee Rhatigan, among others.