An Activision researched and development team demoed new lifelike facial rendering technology to an audience at Game Developer’s Conference.
While the number of facial animations presented are short in length, what they do represent is fairly stunning. Activision converted source material from USC Institute for Creative Technologies into a “70 bones rig” using advanced techniques to deliver realism to the eyes and skin.
When they talk about “bones” they are referring to skeletal animation, which is used in all modern games and movies. Back in the day, vertex animation required animators to animate polygon by polygon, which is not exactly ideal when you’re dealing with a character composed of millions of polygons. The idea with skeletal animation is that animators set up a “rig” composed of virtual objects called “bones” that control the movement of sets of polygons that are actually rendered on screen. Bones can influence each other to create complex animations quickly.
Activision says this facial animation demo can be rendered at 120 frame per second on a Nvidia GeForce 680. But this demo is of a floating, talking bald head, so it does not exactly represent a true-to-life video game scene. This Activision technology was based upon a demo called Face Works which was recently shown during Nvidia’s GPU Technology Conference.
Nvidia says their demo used about half the processing power of a Nvidia GeForce GTX Titan, which is the most powerful $1000 single card video card you can purchase at the moment. When Sony unveiled the PlayStation 4 they also showed a demo of an extremely detailed face of an old man.
So is this technology likely to see the light of day in the PlayStation 4 or Xbox 720? Typically, when video games switch over to realtime rendered cinematic scenes they tend to use techniques and levels of detail that are not feasible during most gameplay sections. So it’s possible that some video games may choose to use such technology when the camera is extremely close to character’s faces.
What do you think about this new lifelike facial rendering technology?