That's the holy grail of CGI? As computer effects teams manage to create ever more photorealistic marvels, one important goal is yet unmet by digital animators: realistic human beings. We’ve all heard the predictions—one day digital wizards will be able to resurrect Humphrey Bogart, Marilyn Monroe, and Rudolph Valentino and have them starring in films again. But even with all the great leaps forward in CG technology, this day still seems very far off, simply because something about computer-generated characters—for the moment, at least—just doesn’t feel real. In fact, some would argue that the more realistic they get, the more downright spooky they become. Welcome to what’s been termed “the uncanny valley.”

Sure, digital magic has helped to resurrect Marlon Brando (in Superman Returns) and Laurence Olivier (in Sky Captain and the World of Tomorrow), but that’s not what we’re talking about here. Those two films just took archive footage and adapted it. It was digital tomfoolery—not an attempt to create fully animated CG versions of those stars. And we’re not referring to that heinous trend for CG stunt doubles, when actors suddenly turn into gravity-defying bendy toys. Think instead of the faux-human characters in Final Fantasy: The Spirits Within. That was back in 2001, and the publicity machine claimed that the film boasted humans so realistic, we’d be amazed. Freaked-out would be nearer the truth: The effects guys may have given the characters pores and moles and hair and imperfections so that their skin could be scrutinized by the closest of close-ups on virtual cameras, but they still looked far from real. In trying to ape reality so closely, the filmmakers actually created something that was unsettling and repellent. Some critics went as far as to liken these virtual actors to zombies. The National Union of Zombie Extras could well have sued.

Five years and countless technological breakthroughs later, director Robert Zemeckis fared little better with The Polar Express and his motion-captured version of Tom Hanks. The zombie-like animated Hanks had a distinctly creepy feel, as if someone had skinned him and stuck a soulless robot skeleton inside his epidermis. Or, as Beowulf screenwriter Neil Gaiman phrased it while promoting that movie, “horrid, little rotoscopey ghost people” (which was an odd observation, considering Beowulf was also directed by Zemeckis, using similar motion-capture techniques).

t’s a strange irony that an audience is happy to identify with the intentionally bizarre, cartoony forms of the lead human characters in, say, The Incredibles...but is repelled by characters who try to look too human. This is what’s meant by “the uncanny valley,” and its effects were being researched even before “CGI” was coined as a term. Initially associated with attempts to create human-looking robots, the theory was first constructed as far back as the 1970s, by a Japanese roboticist named Masahiro Mori.

It’s not quite as simple as saying attempts to make robots or CG characters look human don’t work. Masahiro’s research into people’s psychological responses to robots showed that—up to a point—they do. According to his initial studies, the closer artificial beings come to resembling human beings, the more positive our emotional response. But as that resemblance approaches 100 percent, our responses shift sharply to disbelief and even disgust. Responses only recover when total realism is achieved. When these responses were plotted onto a graph, they resulted in a sudden dramatic dip—hence the use of the term “valley”.

Academic Matthew Butler, who’s written a paper comparing Final Fantasy: The Spirits Within with The Incredibles (admit it, you want a job like that), likens it to going to Madame Tussauds to see the waxworks: “For all the detail that goes into the models, you can’t help feeling that something’s not right.”

Now that CG characters are achieving levels of resemblance that fall into this transitional zone, Mori’s uncanny valley presents a problem to games and special effects artists.

And according to research presented at 2007’s Siggraph (the annual international conference on computer graphics), that problem may be even more complex than was first thought. At the show, a panel including Karl MacDorman of Indiana University, Thierry Chaminade of University College London, and King Kong VFX supervisor Joe Letteri gathered to discuss new theories and evidence.

Even the meaning of “uncanniness” itself is in doubt: In his research, MacDorman conducted an experiment screening 17 robot video clips and one human clip using 143 Indonesian participants who had never been exposed to robots. They rated the videos according to a list of 27 emotions and four terms including “eerie,” “creepy,” “strange,” and “human-looking.” MacDorman’s study then mapped local topological relationships showing which emotions were closest to each other. His conclusion was that fear stood out as being a strong predictor of eeriness—a significant finding, as past studies have shown disgust is the main factor.

Research also shows that uncanniness varies with the body part on show. Eyes, mouths, and hands are cited as the most disturbing regions, depending on the subject's age.

Letteri, who worked on the CG character Gollum in The Lord of the Rings movies, provided the industrial perspective on the issues at stake, revealing that while Andy Serkis’ original live-action performance was flawless, if the animation team matched it exactly, the CG character ceased to look believable. To make the character come alive, they had to soften up some aspects, and emphasize others.

In fact, when asked to pick the factor most critical to the phenomenon, every panelist cited variations in movement. MacDorman said jerkiness—the lack of synchronization between speech and gesture—is a major issue with robotics. Letteri felt that perfecting coordinated and overlapping movements was critical, and Chaminade suggested that synchronization of micro-movements needed improving.

So where does that leave CG characters in movies? Some filmmakers feel that the uncanny valley idea is simply overrated—and that if the storytelling and direction are good enough, then they negate the feeling of uncanniness. In an interview with 3D World magazine, Beowulf’s senior visual effects supervisor Jerome Chen didn’t believe that reaching for a degree of realism put the movie in danger of entering uncanny valley territory. “I’ve read a lot about the uncanny valley, but I don’t know that I really believe in the idea,” he says. “I know there was some criticism about The Polar Express, but I think if the results were that repulsive, then nobody would have gone to see it. And really, it’s a theory related to robotics. In a storytelling environment, it’s just not applicable. If you’re drawn in by the story, that’s all that matters.”

So, the question remains—will filmmakers ever produce the perfect synthetic human? Well, you can always trust good old James Cameron (Terminator 2, Titanic) to take up the challenge. In the director’s next opus, Avatar, he aims to solve the problem in the most unexpected way. Avatar is one of the new wave of 3D movies (you know, the sort you watch in funny glasses), which Cameron believes will help with digital verisimilitude: “Avatar will make people truly experience something,” he says. “One more layer of the suspension of disbelief will be removed. All the syn-thespians are photorealistic. Now that we’ve achieved it, we discovered that CG characters look more real in 3D than in 2D. Your brain is cued it’s a real thing, not a picture, and discounting part of [the] image that makes it look fake. Avatar is the single most complex piece of filmmaking ever made. We have 1,600 shots for a 2.5-hour movie. It’s not with a single CGI character, like King Kong or Gollum. We have hundreds of photo-realistic CG characters.” If anyone can pull it off, it’s Cameron.