Mickey Mouse is one of the ancestors of artificial life. Mickey,
now 66 years old, will soon have to face the digital era. In one of the
permanent "temporary" buildings on the backlot of Disney's Glendale
studios, his trustees were cautiously planning ways to automate
animated characters and backgrounds. I spoke to Bob Lambert, director of
new technologies for the Disney animators.
The first thing Bob Lambert made clear to me was that Disney was in no
hurry to completely automate animation. Animation was a handcraft, an
art. Disney Inc.'s great fortune was sealed in this craft, and their
crown jewels -- Mickey Mouse and pals -- were perceived by their customers as
exemplary works of art. If computer animation meant anything like the
wooden robots kids see on Saturday morning cartoons then Disney wanted
no part of it. Lambert: "We don't need people saying, 'Oh damn, there
goes another handcrafted art down the computer hole.'"
Then there was the problem of the artists themselves. Said Lambert,
"Look, we have 400 ladies in white smocks who have been painting Mickey
for 30 years. We can't change suddenly."
The second thing Lambert wanted to make clear was that Disney had
already been using some automated animation in their legendary films
since 1990. Gradually they were digitizing their worlds. Their animators
had gotten the message that those who didn't transfer their artists'
intelligence from their heads into an almost living simulation would
soon be dinosaurs of another kind. "To be honest," said Lambert, "by
1992 our animators were clamoring to use computers."
The giant clockwork in Disney's The Great Mouse Detective was a
computer-generated model of a clock that hand-drawn characters ran over.
In Rescuers Down Under, Oliver the Albatross dove down through a virtual
New York City, a completely computer-generated environment grown from a
large database of New York buildings compiled by a large contractor for
commercial reasons. And in The Little Mermaid, Ariel swam through
clusters of fish whose schooling was simulated, seaweed that swayed
autonomously, and bubbles that percolated with physics. However, with a
nod to the 400 ladies in white, each frame of these computer-generated
background scenes was printed out on fine painting paper and
hand-colored to match the rest of the movie.
Beauty and the Beast was Disney's first movie to use "paperless
animation," at least in one scene. The ballroom dance at the end of the
film was composed and rendered digitally, except for the hand-drawn
characters of the Beast and Belle. The shift in the movie between the
real cartoon and the faked cartoon was just slightly noticeable to my
eye. The discontinuity protruded not because it was less graceful than
the hand animation, but because it was better -- because it looked more
photographic than the cartoon.
The first Disney character to be completely paperless was the flying
(walking, pointing, jumping) carpet in Aladdin. To make it, the form of
a Persian carpet was rendered on a computer screen. The animator bent it
into its poses by moving a cursor, and then the computer filled out the
"between" frames. The digitized carpet action was then added into the
digitized version of the rest of the hand-drawn movie. Lion King,
Disney's latest animation, has several animals that are
computer-generated in the manner of the Jurassic dinosaurs, including
some animals with semi-autonomous herding and flocking behaviors. Disney
is now working on their first completely digital animation, to be
released in late 1994. It will feature the work of an ex-Disney
animator, John Lassiter. Almost the entire computer animation will be
done at Pixar, a small innovative studio located in a remodeled business
park in Richmond Point, California.
I stopped by Pixar to see what kind of artificial life they were
hatching. Pixar has made four award-winning short computer animations
done by Lassiter. Lassiter likes to animate normally inanimate objects -- a
bicycle, a toy, a lamp, or knick-knacks on a shelf. Although Pixar films
are considered state-of-the-art computer animations in computer graphic
circles, the animation part is mostly handcrafted. Instead of drawing
with a pencil, Lassiter uses a cursor to modify his computer-rendered
3-D objects. If he wants his toy soldier character to be depressed he
goes into his figure's happy face on the computer screen and drags the
toon's mouth into a droop. After testing the expression he may decide
the toy soldier's eyebrows really shouldn't droop so fast, or maybe its
eyes bat too slowly. So by cursor-dragging he alters the computer form.
"I don't know how else to tell it what to do, such as making its mouth
like this," says Lassiter, forming an O with his mouth in mock surprise,
"that would be any faster or better than doing it myself."
I hear more of this communication problem from Ralph Guggenheim,
production director at Pixar: "Most hand animators believe that what
Pixar does is feed scripts into a computer and out comes a film. That's
why we were once barred from animation festivals. But if we were to
really do that, we could not create great stories....The chief
day-to-day problem we have at Pixar is that computer animation reverses
the animation process. It asks animators to describe before they animate
what it is they want to animate!"
Animators, true artists, are like writers in that they don't know what
they want to say until they hear themselves say it. Guggenheim
reiterates, "Animators can't know a character until they animate it.
They will tell you that it is very slow going in the beginning of a
story because they are becoming familiar with their character. Then it
starts speeding up as they become more intimate with it. As they get to
the halfway point of the film, now they know the character well and they
are screaming through the frames."
In the short animation Tin Toy, a plume on the toy soldier's hat shakes
naturally when he bobs his head. That effect was achieved with virtual
physics, or what the animators call "lag, drag, and wiggle." When the
base of the plume moved, the rest of the feather acted as if it were a
spring pendulum -- a fairly standard physics equation. The exact way the
plume quivered was unpredicted and quite realistic because the plume was
obeying the physics of shaking. But the face of the toy soldier was
still manipulated entirely by an experienced human animator. The
animator is a surrogate actor. He acts out a character by drawing it.
Every animator's desk has a mirror on it that the animator uses to draw
his own exaggerated facial expressions.
I asked the artists at Pixar if they can at least imagine an autonomous
computer character -- you feed in a rough script and out comes a digital
Daffy Duck doing his mischief. There was uniform grave denial and
shaking of heads. "If animating a believable character was as easy as
feeding a script into a computer, then there would be no bad actors in
the world," said Guggenheim. "But we know that not all actors are great.
You see tons of Elvis or Marilyn Monroe impersonators all the time. Why
aren't we fooled? Because the impersonator has a complex job knowing
when to twitch the right side of his mouth or how to hold a microphone.
If a human actor has difficulty doing that, how will a computer script
do it?"
The question they are asking is one of control. It turns out that the
special effects and animation business is an industry of control freaks.
They feel that the subtleties of acting are so minute that only a human
overseer can channel the choices of a digital or drawn character. They
are right.
But tomorrow, they won't be. If computer power continues to increase as
it has, within five years we'll see a character created by releasing
synthetic behavior into a synthetic body star in a film.
The Jurassic Park dinos made it very clear how nearly perfect synthetic
body representations are today. The flesh of the dinos was visually
indistinguishable from what we'd expect a filmed dinosaur to be. A
number of digital effects laboratories are compiling the components of a
believable digital human actor right now. One lab specializes in
creating perfect digital human hair, another concentrates on getting the
hands right, and another on generating facial expressions. Already,
digital characters are inserted into Hollywood films (without anyone
noticing) when a synthetic scene demands people moving in the distance.
Realistic clothing that drapes and folds naturally is still a challenge;
done imperfectly it gives the virtual person a clunky feel. But at the
start, digital characters will be used for dangerous stunts, or worked
into composite scenes -- but only in long shots, or in crowds, rather than
in the full attention of a close-up. An entirely convincing virtual
human form is tricky, but close at hand.
What is not very close at hand is simulating convincing human action.
Especially out of reach is convincing facial behavior. The final
frontier, the graphics experts say, is the human expression. A quest for
control of a human face is now a minor crusade.
continue...
|