Monday, November 15, 2004

The Polygon Fallacy and Repellent Realism

Alvy Ray Smith, one of Pixar’s founders, is often quoted as stating that a computer would have to process 80 million polygons a second before it would be indistinguishable from reality. While I have no idea how he settled his figure, I don’t doubt have any reason to doubt it. It’s easy to use this figure improperly however, in a type of “necessary but not sufficient” mistake we might call the “polygon fallacy.”

As a glaring example of what I mean, virtual reality pioneer Mark Pesce, in his book The Playful World: How Technology is Transforming Our Imagination, tells the story of Sony’s PlayStation 2 as a quest to come closer to the 80 million figure. The end product fell a good 60 million polygons per second short of this holy grail, despite being 50 times faster than the original PlayStation. (There are many ways to measure such things. I’m just using Pesce’s numbers because I have them in front of me.) But the PlayStation 3 is supposed to be 1,000 times faster than the PS2, he says, going on to conclude that, if this is so, then “the images it generates will be completely indistinguishable from the real world, as complex and as rich as anything you might encounter in real life.”

Whoah, partner! Where is this complexity and richness supposed to come from? Someone or something has to create that level of complexity before it can enlist a machine to render it. It’s just not practical to do this today, and there are in fact good reasons to not jump at every opportunity to increase realism. The better game design firms, such as Blizzard, don’t load their gameplay or cut-scenes with the motion-captured images of Hollywood actors, although it would be cheaper and easier for them to do so. In the context of most game universes, which are necessarily—even intentionally—much less realistic, such tricks draw too much attention to themselves. They make the necessary and voluntary suspension of disbelief more difficult for the gamer to maintain.

This is a lesson the creators of the new movie The Polar Express might have wanted to take into account, as they captured an unprecedented amount of data from actors’ faces (mostly Tom Hanks) to actuate various characters in the film. This technique produces the most realistic animations of people ever, but at what cost? I haven’t seen the movie yet, but if the reviews are to be believed, then the effects wizards have managed to hit that known sweet spot of creepiness where the mind is simultaneously drawn inward by realism and repulsed by subtle imperfections. Pixar and others have deliberately kept their characters more cartoony for this reason. Is it not better to harness reality only inasmuch as it draws the audience into the story? Is it not wiser to avoid state-of-the-art gimmicks that merely draw attention to the medium?

That said, I do expect the richness and complexity of game environments to increase over time. As textures, shading, and physics improve, motion captured actors will not seem so out of place. Better tools for digitizing the properties of actual materials and the contours of actual locations will also help bridge the already narrowing gap between production techniques effective on film and in games. But for the foreseeable future, gaming environments are just not going to be mistaken for television programs, no matter how many billion polygons come screaming at you.


Post a Comment

<< Home