UC Berkeley researchers developed a technology to alter dance videos, transferring video to potentially make amateur dance moves seem more advanced.
They published their research in a paper entitled “Everybody Dance Now,” which explains that the technology can transfer video of one person’s movements onto video of another person’s movements. According to the published paper, the technology attempts to employ what is known as “do as I do” motion transfer.
“With our framework, we create a variety of videos, enabling untrained amateurs to spin and twirl like ballerinas, perform martial arts kicks or dance as vibrantly as pop stars,” the research paper said.
The team of researchers included UC Berkeley alumna Caroline Chan, who majored in computer science and applied math, Tinghui Zhou and Shiry Ginosar, who are doctoral candidates in the UC Berkeley computer science department and Alexei Efros, campus professor of electrical engineering and computer science.
“We actually already had some people contact us who can’t dance for various reasons — people with disabilities, for example. We’re looking into ways to film them and maybe animate them,” Ginosar said, who explained that the team intends to use the technology to help people.
When it is ready, the technology will likely be released as code on GitHub, according to Ginosar. She said the code is not “perfect,” and there are still some glitches that need to be worked on before it can be accessed by the public.
Currently, the technology is only available to a few individuals and is difficult to operate unless one knows how to navigate the complex code. Ginosar said they are still thinking about how to make it accessible to the public and in what direction they should take their work.
Ginosar added that the code could be co-opted for negative purposes, such as creating false videos.
“This is generated fake data. You could use it for all kinds of bad, making people move in ways they don’t want,” Ginosar said. “Here, you add one dimension, which is motion. You could make someone believe they did something they didn’t do.”
Although the technology could create videos that display falsehoods, Ginosar said she is “very excited” about the project, which works better than the team expected. She added that she does not believe their project goes against professional dancers, because the videos they create are clearly different from reality.
“Every time there is a new medium for people with art, on one hand people are hesitant,” Ginosar said. “If we make a video of you dancing, it’s not like you’re actually dancing. There’s a difference between moving in real life and moving on a screen. It’s like having friends on Facebook versus having friends in real life.”