L.A. Noire ism without a doubt, still one of the leading hallmarks of facial capture industry. Team Bondi built stunning proprietary technology to capture their actors every suitable expression. The results of their labor are detailed down to the slightest creases in the performers foreheads. But even though the faces of L.A. Noire are remarkably detailed, that’s a fact that is often contrasted against models with stiff bodies and an overall feeling like a lively face has been pasted over a head like some horrific Face/Off scenario. It can lend the game an unsettling quality. The reason for this is that L.A. Noire rests firmly in the uncanny valley.

In 1970, a Japanese robotics expert, Masahiro Mori, coined the phrase Bukimi No Tani, or valley of eeriness, to describe a problem he was running into in the creation of his robots. What Mori found was that when he gave his creations some human characteristics, people thought they were charming, but as he gave them expressions and brought them even closer to lifelike he found that they became off putting. Like Team Bondi’s technology today, the techniques and aesthetic design he was pioneering was impressive, but unsettling.
Eight years later a Polish writer by the name of Jasia Reichardt coined the most popular English translation of Mori’s descriptor for his conundrum: The uncanny valley.
On one side of the valley you have robots or images devoid of any human qualities whatsoever. When you start to give them some stylized human features, they become more appealing. That is until they become a little too human. When something approaches bearing a photorealstic resemblance to humanity, but is still marred by subtle idiosyncrasies not normally exhibited in real human faces, we find them creepy. It’s not until those imperfections are ironed out and the face is perfectly realistic, both in detail and movement, that we go back to finding it appealing.
That dip between cute stylized simulacrums and perfect renditions of human faces is the uncanny valley.
We as a species have evolved to pay particular attention to faces, and we’re quite good at knowing when things feel just slightly off. We’re also more inclined to give more leeway to stylized caricatures, and less inclined to find subtle deviations in what a human face looks like palatable if the face in question is more life-like.
The reason we know this is because the uncanny valley has been the topic of research in the field of behavioral psychology. In particular, one experiment rendered 18 versions of a CGI human face based on an actual face. These 18 versions ranged from very simple to highly detailed models with a large number of polygons in order to achieve photorealism. Participants in the experiment were asked to arrange facial details to either make the 3D faces more palatable, or more disturbing. It was comparatively difficult for participants to make something unpalatable by tweaking, for instance, the amount of separation between the eyes using the lower detailed models. On the other hand, making the eyes of the photorealistic models just slightly too far apart caused them to be seen as uncanny.
We’re incredibly perceptive when it comes to spotting when a face looks wrong, as if it doesn’t belong to an actual human. Take for instance this scene in Blade: Trinity. I bet something looks pretty off to you about it. That’s because Wesley Snipes isn’t actually blinking. Due to turmoil on the set, Snipes refused to open his eyes during the shooting of this scene, so the director had the production team add CGI eyes overlaid on to Snipes’ face. Something that subtle is very obviously wrong to you and I.
When we don’t see something with human qualities that isn’t trying to accurately emulate the human face, we’re more likely to empathize with it and find appeal in it. Pixar characters take great advantage of this to create memorable and appealing characters with advanced CGI that doesn’t even come close to trying to emulate photorealism. They instead lean heavily upon stylized expressions and faces.
It’s not impossible to live on the other side of the valley with photorealistic faces, but it is so easy to fail and slip into the uncanny. In my opinion, Horizon: Zero Dawn, released this year, is the closest games have yet come to bringing lifelike faces out of the valley, but there’s still a fair amount of criticism that the game and its characters have received. We’re a far cry from early Playstation CGI where the disconnect was at its worst, but we aren’t quite there yet with photorealistic graphics. In the mean time, some of the best in the business are working on creating appealling lifelike faces, and there’s always the left-most side of the valley full of charming stylized characters to enjoy as well.
Sources:
1. http://www.psy.gla.ac.uk/~frank/Documents/InSearchUncannyValley.pdf
2. http://www.macdorman.com/kfm/writings/pubs/MacDorman18Dec2008TooRealForComfort.pdf
3. https://www.jove.com/video/4375/perceptual-category-processing-uncanny-valley-hypothesis-dimension
4. https://commons.wikimedia.org/wiki/File:Mori_Uncanny_Valley.svg
5. https://www.livescience.com/20909-robotics-uncanny-valley-translation.html
Leave a Reply