Although the concept of the uncanny valley originated in robotics, it has generated higher levels of serious consideration in the computer graphics industry. Despite being largely unsubstantiated by empirical evidence, the phenomenon is generally accepted in the public mind-space as it pertains to photorealistic human animation and rendering for film and games.
The term was created by Masahiro Mori in a 1970 article for Energy magazine to describe the emotional response curve experienced by humans observing lifelike non-human entities. Mori’s original hypothesis plotted emotional response (positive or negative) or familiarity against increasing levels of human likeness and more or less predicted the following:
As the appearance of a robot (or photorealistic 3D model in our case) is made more human, the observer’s emotional response will become increasingly positive until the likeness reaches a point where slight imperfections become emphasized and the emotional response turns overwhelmingly negative.
In other words, the more realistic a computer generated human looks, the easier it is for our brain (specifically, our facial recognition system) to spot its imperfections. There’s another part of Mori’s hypothesis:
As the appearance of the entity continues to become less distinguishable from a human being, the emotional response becomes positive once more and approaches human-to-human empathy levels.
The phenomenon is most easily represented in chart form. As you can see in the image above, emotional response becomes increasingly positive as human likeness increases, and then drops off sharply into the “uncanny valley,” as described in Mori’s first prediction. The second part of his hypothesis is shown at the far right of the diagram, where the emotional response curve begins to climb until it peaks when human likeness becomes 100 percent accurate.
How Does it Pertain to 3D Computer Graphics?
The uncanny valley phenomenon is an important consideration in the lives of concept artists, production designers, 3D modelers, animators, and render specialists. Over the course of the last decade, there have been numerous attempts at photorealistic human rendering, may of which have failed and ended up somewhere in the depths of the uncanny valley, and a few that have succeeded in climbing (at least partially) the far wall of the canyon.
Here's a small list of some notable films and games that have tried, and at least partially failed to depict near (or complete) human photorealism. The most famous example is probably Rob Zemeckis' The Polar Express, which received almost universal derision for being "creepy."
In our opinion, the worst offender by far is Pixar's 1988 short film, Tiny Toy, which is doubly uncanny because the subject in question is an infant. Of course, they can be forgiven for the fact that it was 1988, when computer graphics were in their — infancy.
- Tiny Toy (1988)
- Final Fantasy: The Spirits Within (2001)
- The Polar Express (2004)
- Beowulf (2007)
- A Christmas Carol (2009)
- Heavy Rain (VG 2010)
- TRON: Legacy (2010)
- L.A. Noir (VG 2011)
- Mars Needs Moms (2011)
The most common criticisms levied against these films include the following:
- Hollow, lifeless eyes - This is most likely a result of inadequate motion tracking. A human eye is never stationary. Also difficult to nail down are the complex specular highlights and reflections on the cornea.
- Waxy, zombie-like skin - With the advent of sub-surface scattering and normal mapping, many problems with skin rendering have been resolved. All living skin is somewhat translucent, and the lack of a good solution for simulating skin translucency plagued many early attempts at photorealism. Normal mapping has meant that modern skin shading-networks can include accurate textures down to the pore level.
- Poor facial animation and lip synching - This issue is more common in games, but nothing deflates a render like improperly synched speech.
Can the Uncanny Valley Be Beaten?
In short, yes. There are two ways the uncanny valley phenomenon can (and has) been defeated:
- Cartoon realism - Animation studios learned long ago that the best way to beat the uncanny valley is to avoid it altogether. Instead of striving for complete photorealism, studios like Pixar, Dreamworks, and Disney have all developed their own unique flavor of cartoon-realism.
Instead of designing realistic human characters, artists at animation studios typically stylize and exaggerate the features on a model that are most difficult to replicate, like eyes, hands, and lips. Animated films like Ratatouille, The Incredibles, and Cloudy With a Chance of Meatballs have lifelike, empathetic human characters that really look quite unrealistic. But they work.
- Innovation - The second way, plain and simple, is with brute-force technological innovation. In fact, with modern sculpting, texturing, and performance capture technology many of the initial boundaries toward complete photorealism have already been resolved. We still haven't seen a fully photo-real CG film that's completely avoided the uncanny valley, but we are at a point where successful human photorealism is possible.
The Curious Case of Benjamin Button managed to traverse the uncanny valley and come out on the other side. The team used a fully computer generated head replacement for Brad Pitt in many shots that required drastic alterations to his appearance. The film was well recieved, the CG was hardly noticed, and the team won a Visual Effects Oscar for their staggering achievement.
Time will tell whether that success can be duplicated for an entire film. It'll probably take ambition on a level that hasn't been attempted since Final Fantasy, but we think it's just a matter of time.