How to create an animated character from your facial expressions in real time

An avatar reproduces your facial expressions in real time. You move, it moves. You smile, it smiles.
November 20, 2012

Faceshift automatically reproduces facial expressions, simplifying facial animation. Credit: Alain Herzog/EPFL)

Faceshift, announced Monday, is software that uses Kinect and similar cameras to read your face and create a moving avatar from it. It could enhance the future of video games and make video chats more fun, says Thibaut Weise, founder of Faceshift, a spinoff of EPFL’s Computer Graphics and Geometry Laboratory.

The software needs only ten minutes to recognize your face as you create basic expressions requested by the program: smile, raise eyebrows, etc. and start reproducing your expressions.

Creating animations faster

Weise’s team has created an algorithm that superimposes 3D depth data on image color and brightness data in one step. They demonstrated that 3D facial movements could be reconstructed in real time without using facial markers or complex scanning hardware.

With the current method of creating animated films and video games, the facial expressions of characters are defined with a program that creates movement of different parts of the face step by step. To simulate anger, for example, you have to knit each eyebrow in two or three clicks, then stretch the mouth down, and so on. Faceshift can reduce the time to make a film by up to 30%, he says.

A free trial version is available.


Faceshift video