But there are other, perhaps under-explored uses for this technology, such as trying to fake a video of you dancing so it looks as though you have Bruno Mars’ moves, instead of Mr Bean’s.
As this video shared by Caroline Chan demonstrates, researchers at UC Berkeley have found a way to copy the movements of a subject’s body in one video, and then generate a new video of a completely different person’s body more or less performing those actions. (For more details, see their paper humorously titled “Everybody Dance Now”.)
It’s an extension of the research and automated image manipulation that can make someone appear to say something they never did, but with the added challenge of "faking" far more movement — particularly if you’re trying to copy the moves of a talented performer such as Bruno Mars.
Digitally morphing and manipulating the movements of someone’s face often requires just subtle changes between the original clip and the modified results. But when copying dance moves, a person’s arms, legs, head and torso could potentially have to move in completely different ways than they did in the sample footage that’s used to train the artificial intelligence.
To make the motion transfer possible, the AI generates simple stick figure representations of the movements of subjects in both the source and target clips. The changes in motion needed to make one person move or dance like another are calculated, and then that data is used to generate new frames of video featuring someone appearing to tear up the dance floor, even if they have two left feet.
As with all of these advancements in AI-assisted image processing, there’s always the potential for this research to be used for nefarious reasons. It’s easy to imagine that one day someone could be framed with faked security camera footage of them appearing to do something illegal.
But in this scenario, I’d argue the pros outweigh the cons, because now I can reprocess all the footage from my wedding so it looks as though I owned the dance floor.