Jamie Antonisse

I've heard several people say that this is the animation that leaps across the uncanny valley into the promised land of believable humanlike CG characters. I agree, this is an impressive demo... but what's most interesting about it is the way it blurs what we mean by a "CG character" in the first place.

In fact, this demo is a little sneaky, a little fuzzy on exactly what it's delivering. Scan the YouTube comments for the linked video (go on, be brave) and you'll see how confused people are about what's actually being presented to them.

We are looking, to my understanding, at a wire mesh matched with incredible detail EXACTLY over the video-captured human face. While this level of matching is a pretty impressive feat for these riggers and animators, what we see is just that... matching, a sort of 3D rotoscope created via what was probably a painstaking trial and error.

Scroll down to the bottom of this linked page for a less popular demo that shows what happens when you remove the direct connection. I'm no longer "convinced" in the same visceral way (sorry, Image Metrics), but at least I get the link between the controllers and the actor. Once again, they made sure to match the original video source's angle and position exactly so that they could be as precise as possible in their simulation.

While this second video shows the potential applications of this technology much more clearly (and brings us back to planet Earth in terms of our expectations), it's not the PR centerpiece. It's the "Emily on Emily" video that's being shown off, because it speaks so directly to our culture's simulation fetish. Look, it's so real that it's... exactly superimposed over reality. Cool, I guess? But is anyone else reminded, in a ridiculously literal way, of the Emperor's New Clothes?

What's undeniably impressive here (though still, in wireframe, a little creepy) is the base facial model, captured from a light stage. Slowly I imagine we'll get better and better techniques for this, and move closer and closer to modeling real performances in full 3D. In gradual steps, this whole "uncanny valley" thing will become moot, and we'll find solutions that approximate the 3D and 4D photograph (which has, in its fictional form, become one of the cornerstones of my thesis).

Does this mean 3D animation will go away? Of course not... 2D animation continues alongside video and painting continues alongside photography. We'll just get back to the proper business of making characters and worlds that are expressive and evocative on their own terms, instead of trying to "wow" people with close-but-no-cigar simulations of the mundane.

I can't help but wonder if that's all the uncanny valley really is: the limitation of resources, and an artistic confusion of the evocative with the real. If you spend all your effort trying to "fool people" into thinking that your characters are actual people (the aim-apparent of Beowulf), the audience will respond in kind. They will play your game, and either be fooled (but not necessarily entertained) or not fooled (and annoyed). Animators using less realistic models, such as the maestros at Pixar, can focus their energies instead on performances that are actually engaging and entertaining to watch.

"Emily", impressive as it is, reminds me that there are many people out there who think simulating reality is an end, not a means. With their functionally transparent CG mask, Image Metrics may have achieved their goal... but it seems like a bit of a hollow win. If you ask me, making something imagined and constructed FEEL truly believable is still a moving target, better approximated by art than by artifice. Until someone steps up with a Replicant to prove me wrong, here is one more attempt at the photoreal that will haunt your dreams.

N.B. - The image above is linked from http://petitinvention.wordpress.com, found via a google search for "see through glass"