Tuesday, July 31, 2007

Motion Cap & The Uncanny Valley

Seems like there's alot of controversy regarding motion capture animaton, similar to the one surrounding rotoscope animation. I personaly dont have a problem with the technique itself, but i find that it is often used with problematic results. Since i dont know all the technical details about going about it, perhaps i shouldnt be too quick to criticize, but i still feel that the same problem keeps coming up.

What bothers me the most about some mocap films is when they model the characters after their voice actors. Think of Tom Hanks in the Polar Express. If they want to character to look so much like Tom Hanks, why not just use Tom Hanks? The same people who made The Polar Express are now releasing a film titled Beowulf based on an old english poem. Its partly written by Neil Gaiman, so imagine the story will be great and all, but i just know the animation in dialogue scenes will be awkward. One of the characters is played by Angelina Jolie... and the character LOOKS JUST LIKE HER.

Its impressive in a way, but at the same time i cant help but notice how much of a waste it is. If you want a character to look real, film him in live action. Even if you have loads of special effects, the characters in dialogue scenes should be live action if "real" is what youre going for. 300 and Sin City are good examples. If the characters in those films were always animated characters, there would be an incredible loss of visual appeal.

It all comes back to this hypothesis i read about a little while ago called The Uncanny Valley (http://en.wikipedia.org/wiki/Uncanny_Valley).

Basically, the more something looks human, the more you can notice the differences (this is amplified for things that move). Therefore if the Jolie character looks almost exactly like Jolie, but not quite, the moment she moves itll feel awkward since she is moving "realistically" through mocap, but doesnt have the perfectly similar appearance. I imagine that eventually we will be able to model characters that are indeed identical to their voice actors, then perhaps this problem wont occur, but even if we get to that point, wouldnt it be alot cheaper to just film the actor in live action?

Last night I saw another mocap film by the name of Renaissance (http://video.movies.go.com/renaissance/).

Its a french film noir/detective/sci-fi film set in Paris. This film at least stylised it's characters alot, however i feel like the animation shouldve been stylised just as much to compensate. What happens is that all the characters move like real people but visibly are not, therefore we get the same awkwardness as in most other mocap efforts. That being said, the film has some very beautiful sequences and is well worth seeing. The sets/layouts/backgrounds are particularly breathtaking. The story is a pretty typical detective story, but fun nonetheless.

As a final note, ive noticed that motion capture is particularly awkward with children. The children in Renaissance felt very odd, so did the ones in Final Fantasy: Advent Children. I cant say i know what it is exaclty, but there's definatley a problem with motion capture animation of children. Even the kids in Moster House felt odd to me, though perhaps less due to the bobble head factor.


Chris MacDonald said...

dude I recently saw Renaissance and agree with you pretty much 100%. I do love how they stylized everything black and white, but most of the time it just looked like a video game, at least when it came to physical movement. I think the issue is speed. Mocap probly cant capture at the same speed film does...That or with all the 3D complexities they put over top the mo-cap...something is lost in the process.
Either way it was an interesting film adn worth watching

Matías Hannecke said...

Beowulf looks so incredibly retarded.. the just want to impress people with new and exciting technology. It'll just end up looking like sloppy awkward live action film. Why the hell do people want live action and animation to be the same? *coughs*..retarded.

PO said...


Hey, that's some pretty interesting stuff you're talking about there. However, theres a big point your missing. The issue of visual awkwardness does come from the uncanny valley thing but motion capture is no to blame. The data that is taken out of motion capture is a raw kind of data. It means that it cant be applied to a 3d model without any work. What happens is that all the movements are recorded individually and then put together by the animation team of the studio. The problem here is that there are very few good animators who can portray the depth of human movement correctly and the ones that can are hired at extremely high price. Most studios think it irrelevant as long as the visual (the actual image) looks good to have perfect animation.

On another point, I think that it can be really interesting to have remodelled actors. The possibilities for action and previously impossible scenes becomes enormous.... Just wait till you see the footage of the short movie my brother directed (technical director).... it kicks butts...
For me seeing a close up of the face of a girl falling from a 40 stories high building, all this at very slow motion is kind of priceless.....
that's what I had to say..... and I'll show you the footage when you get back here next week