A Graphics Breakthrough Makes Perfect CGI Skin

By Andrew Liszewski on at

You might not understand all of the technical details behind the computer graphics research being revealed at Siggraph 2015 this week, but come next year when the CG characters in movies and video games start to look indistinguishable from real humans, you’ll know who to thank.

In this case it’s a team of researchers from the USC Institute for Creative Technologies, and Imperial College London that jointly developed a new method for capturingthe impossibly subtle deformations in human skin, and translating them to a computer-generated character.

Motion capture, in which an actor’s physical performance is recorded and translated to a CG character they’re portraying, might seem like a shortcut when it comes to animation. But the technique is able to reproduce incredibly subtle performance gestures and motions an animator might miss.

Motion capture is also great for facial performances. Though still, capturing every last wrinkle and deformation of human skin is basically impossible for talented CG animators and sophisticated video capture systems alike. Those infinitely subtle ways the skin moves on a human face allows us to spot the differences between a real-life humans, and their CG counterpart — at least for now.

A Graphics Breakthrough Makes Perfect CGI Skin

The team of researchers has come up with a new way to capture the tiny details on the surface of various skin patches on an actor’s face at a resolution of 10 microns as they’re being stretched and deformed by a specially designed rig. At that resolution the exact deformations of even individual skin pores are captured, and using custom software, the captured data can be mapped to the artificial skin of a CG character, making the emotions of the face so realistic you can’t tell human from computer human.

Eventually the researchers hope to design a model that allows the system to perfectly simulate exactly how human skin deforms based on a person’s race, age, sex, and emotions, without requiring actors to be subjected to that face deforming contraption. That also means that one day the terrified looks on the faces of citizens in future editions of Grand Theft Auto will be so realistic, they might actually haunt you after you run them down. [USC Institute for Creative Technologies]