Watch: With this AI video tech, anybody can dance like Bruno Mars

Anybody can look like a pro on the dancefloor with a little bit of help from deepfake technology. A recent trend to appear on the internet is ‘deepfake' -- deep learning technology used to superimpose existing images or video onto other sources. This type of AI can be used to make politicians look as though they're saying something completely different or it could make someone with two left feet appear to dance like Beyoncé.

“Everybody Dance Now” is the title of the paper by researchers at University of California Berkeley,  who are using their AI framework to give untrained amateurs the chance to “to spin and twirl like ballerinas, perform martial arts kicks or dance as vibrantly as pop stars.”

The idea is to transfer motion between human subjects, a complex task made up of three stages that the team has broken down and explained as a "simple method for 'do as I do' motion transfer" -- “pose detection, global pose normalization, and mapping from normalized pose stick figures to the target subject.” Simply put, the dancer's movements are mapped out onto a computerized stick figure who acts somewhat as a middle man, then those motions are transferred to the ‘amateur' target.

The researchers noted that some issues needed to be addressed, such as movements that needed to be smoothed out while trying to maintain minimal blur. Furthermore, the team found that their pose representation didn't encode information about clothing, so loose shirts wouldn't work as well as a tight t-shirt. One of the researchers posted a video on YouTube to demonstrate the process and capabilities -- one of the clips using Bruno Mars'