Skip to main content

This face filter technology can turn you into a statue or an oil painting

This face filter technology can turn you into a statue or an oil painting


Are we looking at a face filter face off?

Share this story

If you buy something from a Verge link, Vox Media may earn a commission. See our ethics statement.

A team of researchers led by a professor at the Czech Technical University in Prague has created an AI-powered algorithm that can animate faces in the style of another photograph. In other words, a bidding war between Snapchat and Facebook is probably about to begin.

The short sample video posted by Professor Daniel Sýkora is reminiscent of recent projects like this one, where researchers were able to manipulate video of people speaking in a way that made it seem like they said something completely different. Another research team recently cracked the ability to transfer styles from one photo to another, allowing them to to take a picture of a skyline at sunset and essentially overlay an even more outrageous sunset on top of it, for example.

Apparently more flawless than Snapchat or Instagram

Sýkora’s software does a little bit of both, but with a particular focus on human faces. It’s able to take a picture of an oil painting on canvas or a bronze statue, and transfer those styles to videos of people speaking. The results are spellbinding. It’s not all that far off visually from what apps like Snapchat, Facebook, and Instagram offer (or even what we know they’re working on), but the execution is far more flawless.

As for how exactly Sýkora’s team is pulling it off, it’s unclear. There’s clearly some type of machine learning-powered tricks at work here, though. The software is capable of identifying faces and separating the foreground and background of photos in much the same way modern computer vision algorithms can.

The team’s paper on the project has apparently been accepted by Transactions on Graphics, one of the many journals published by the Association for Computing Machinery. But that paper won’t be published until later this summer, when Sýkora’s team presents their findings at the ACM’s SIGGRAPH conference.