November 06, 2007

The "uncanny valley"

Reviews of the $165 million animated feature The Polar Express made using extensive use of motion capture technology as well as digital smoothing and texturing, has resurrected Japanese roboticist Masahiro Mori's warnings about the dangers of creating artificial life forms that are "almost but just not quite human." What Mori discovered, explain Clive Thompson, was that "the more humanlike his robots became, the more people were attracted to them, but only up to a point."

When a robot becomes 99 percent lifelike—so close that it's almost real—we focus on the missing one percent. We notice the slightly slack skin, the absence of a truly human glitter in the eyes. The once-cute robot now looks like an animated corpse. Our warm feelings, which had been rising the more vivid the robot became, abruptly plunge downward. Mori called this plunge "the uncanny valley," the paradoxical point at which a simulation of life becomes so good it's bad.

This isn't as big a problem in Final Fantasy, the first feature film to replace an entire cast with 3D animated androids. Back in 2001 disbelief still required suspending, which, frankly, was to its credit. Final Fantasy teetered on the edge of the uncanny valley. But with four more years of technology to work with (a lifetime in microchip terms), The Polar Express clings precariously to the edge.

On the negative side, writes Jason Silverman in Wired,

the characters are, unintentionally, more creepy than sweet. A girl on the train has an odd immobility in her lower face—she looks a bit deranged. And the eyes and face of a young boy are robotic and blank—he reminded me of David in A.I., but far less expressive.

Agrees Eric Snider, "People's faces look especially creepy, like robots trying to approximate human behavior." It will be interesting to see how well James Cameron executes his live-action version of Battle Angel. Alita, he says, will be digitally animated. But Alita, like the Terminator (and more particularly, T-1000 and T-X), is an android, so it just might work. We won't expect her to look entirely human.

According to Mori, artists and designers "should not strive overly hard to duplicate human appearance," lest some seemingly minor flaw drop the hapless android or cyborg into the uncanny valley." The final product, Dave Bryant points out, "should be visibly artificial, [yet] smart and stylish in appearance . . . [an aesthetic approach that] can be seen . . . in manga and anime."

The effect of this is apparent in Innocence, the sequel to Ghost in the Shell, which mixes "traditional" and digital animation. At least to my eye, the simpler, hand-drawn human characters look far more "real" than the digitally-created people in state-of-the-art productions such as Shrek.

The Pixar approach so far has to stay away from the human, or to keep them cartoony (the wonderful Incredibles, for example). The feature film production of Appleseed takes that retrogression a step further. It makes them look like anime characters again.

Hair, for example, is difficult to render digitally, so in Appleseed they don't try. Hair is stylized the same chunky way it is in anime. Shading as well uses a limited palette. The result are low-rez renditions that appear hand-drawn in some scenes and like sophisticated marionettes in others.

Just as important is how characters move. Watching Appleseed, the eye is drawn particularly to the depictions of physical movement. The brain then tries to interpret the human it recognizes from the motion-capture in the context of something that doesn't look "human." The result is both unsettling and weirdly compelling.

On the other hand, the digitally-animated Tachikoma robots in Ghost in the Shell: Stand Alone Complex are so tremendously appealing precisely because they don't look "human" in the slightest, and yet the motion capture lends them an almost "organic" presence.

We're not "there" yet, warns Clive Thompson. "The more they plug away at it—the more high-resolution their human characters become—the deeper they'll trudge into the Uncanny Valley." Summarizes James Surowiecki, speaking to the same problem with video games,

After a certain amount of graphical improvement, it actually gets harder to suspend your disbelief and truly immerse yourself. The closer a game gets to resembling the real world, the ways in which it's different become more obvious, and the more psychologically jarring those differences become.

Besides, if we do get "there"—if you can't tell the difference between "real life" and "art"—then what's the art for? The Louvre would replace the Mona Lisa with a (digitally scanned, of course) photograph. And, frankly, after watching Appleseed, you don't think, gee, if they only had a few more million to render the hair and skin textures correctly. You think, gee, if only they hadn't relied on a plot device (big computer takes over world, messes everything up) that was old twenty years ago.

Manga artists instead employ a minimalistic, representational approach that surprises you with its simplicity when you observe it closely. Compare these two frames from Lament of the Lamb by Kei Toume. Only the slightest aleration of line vividly depicts Chizuna's change of mood. But to know what caused that change of mood requires you to read the text.

To borrow from Marshall McLuhan's terminology, the manga is "hotter" than the comic book. Like prose, it requires the reader to actively engage in the narrative structure, not just look at the pictures, in order to follow the story. You rarely see in manga the full-color, "artistic" if idealized human forms that you do in American superhero comics.

Inanimate objects such as mecha typically receive far more attention in design than do their human pilots, who define their character according to their human behaviors, not their human appearance. Rather, characters are physically differentiated by what may be called "meta detail," often highly symbolic in nature. In this respect, the stylistics in manga bear more than a passing resemblance to those of the traditional Japanese theater, Kabuki.

Explains Donald Shivley, the onnagata (a female impersonator in Kabuki drama) "singled out the most essential traits of a woman's gestures and speech and gave to these a special emphasis in much the same way that puppets exaggerate human gestures to appear alive."

Consider the use of shading and color when it comes to hair and eye color. For example, Alisa in Fruits Basket is drawn as a blonde to indicate that she is half-Caucasian. Needless to say, an genetic improbability.

Likewise, Frederick Schodt points out that the color inserts in manga often depict even native Japanese characters with "distinctly blonde hair and blue eyes," illustrations that are clearly not intended to be taken literally.

The quintessential example is Ranma, whose female half is depicted with red hair, as contrasted to the black-haired male. Liana Sharer notes that red hair tags a character (as it does in western culture) as "feisty," indicates spirit possession (Ranma's split personality is the result of an ancient Chinese curse). And, especially with male protagonists, "emphasize[s] their traditional nature." Antonia Levi agrees that hair color "not only serves to differentiate the two personas, but also gives a color cue as to which is the true, the proper Ranma."

However, this abstract, minimalistic approach means that if you are unfamiliar with the metaphors employed, it can sometimes get hard to tell characters apart. You have to pay attention both to the text and the subtext. It is a more artistic and "literary" technique that pushes manga closer to prose and pulls these "traditional" 2-D anime characters further away from the edge of the "uncanny valley."

Related posts

The uncanny abyss

Labels: , , , ,