Issue: Volume: 23 Issue: 12 (December 2000)

The Sleigh and the Rocks and the Whos and their Socks

This year, we've seen realistic storms at sea and humans created with computer graphics, science fiction effects and animated films, digital dinosaurs, and 3D cartoon characters in live-action backgrounds. What better way to end the year than with a fantastical, magical, storybook world created with computer graphics? And what better story to tell than Dr. Seuss's much-loved How the Grinch Stole Christmas?

The Universal Pictures' release of the film stars Jim Carrey as the old Grinch who steals all the presents in the town of Who-ville only to learn, with the help of Cindy-Lou Who, that the true spirit of Christmas doesn't come from a store. Ron Howard directed the movie, which also stars Taylor Momsen as Cindy-Lou Who and Christine Baranski as Martha-May Whovier. Digital Domain (Venice, CA), provided most of the visual effects, using computer graphics to help create the whimsical world of Who-ville and, as Dr. Seuss would say, "a little bit more."
In shots where you clearly see the Grinch (Jim Carrey) in the sleigh, the sleigh is real. Otherwise, it's digital, as are the trees and snow in the background.
© Universal Pictures. Images courtesy Digital Domain.

"It's the biggest movie we've done," says Kevin Mack, visual effects supervisor. "We've got a little bit of everything in this one." The studio, which won an Academy Award for Titanic, is known for creating invisible visual effects-photorealistic effects that blend seamlessly into live-action movies such as Apollo 13 and True Lies, as well as surreal effects for Fifth Element and What Dreams May Come. For How the Grinch Stole Christmas, Digital Domain applied the technology and techniques developed to create photorealistic real worlds to creating instead what the people at the studio began to think of as "a real world, just not our world."
Wearing makeup created by Rick Baker and, for this shot, a grin extended with a morph at Digital Domain, Jim Carrey becomes the Grinch.

By the time the effects crew finished, they had modeled entire mountain ranges, painted skies and filled them with 3D volumetric clouds, sprinkled the town with digital Whos, animated a wild sleigh ride down the snowy, rocky all-digital Mt. Crumpit, created tons of snow, rendered thousands of trees, made bugs and bananas, fire and icicles, a girl and a dog, and much, much more. The crew used Alias|Wavefront's (Toronto) Maya primarily for the character pipeline and to create terrain, buildings, the sleigh, and other models; Side Effects' (Toronto) Houdini was used primarily for the effects pipeline and for procedural animation, although both programs were used interchangeably. Mental Images' (Berlin) Mental Ray and Pixar Animation Studio's (Pt. Richmond, CA) PR RenderMan handled the rendering and Interactive Effects' (Santa Monica, CA) Amazon Paint created skies and textures. In addition, the crew relied on in-house tools, notably the compositing software, Nuke, and on conduits that link all the programs.

The film has CG effects throughout-even scenes that seem to be entirely live action will have CG snow falling. Of the 354 shots created at Digital Domain, about 30 were made entirely with computer graphics. "There are probably another 50 that are 80 to 90 percent CG and another 50 that are 70 percent or more," Mack adds. Most of the all-CG shots take place outdoors, often in full daylight.

"Our biggest challenge was creating the world the Whos live in," Mack says. "The idea was to create a real sense of place so that over the course of the film the geography would become clear; the audience would sense which way is north."
To make a "Pee Wee" march out of a tuba, Digital Domain used Maya for modeling and animation, and Mental Image for rendering.

Production designer Michael Corenblith designed the Who-ville buildings; special effects make-up artist Rick Baker turned Jim Carrey into the Grinch using green makeup, prosthetics, and yak hair, and turned the other actors into Whos. Their designs all reflect a Seussian style: The Whos look like people but not quite; and there are no straight lines in Who-ville. Physical sets were built for the town square, the Grinch's cave, the sleigh, the very top of Mt. Crumpit, and numerous interiors. Virtual sets were built for the rest of the environment. Digital Domain extended the Who-ville and Mt. Crumpit sets using Corenblith's designs, cloned the practical sleigh, and designed and built Who valley, the mountain ranges, and the skies.

"We really struggled to come up with a style of sky and clouds and mountains that would simultaneously be Seussian and have a sort of cartoon flair yet be completely realistic and three dimensional so that you could believe that you were there," says Mack. "The Dr. Seuss style is so identifiable that you think it would be easily quantifiable, but as it turns out, it's rather elusive."

"Before we did anything else, Kev in Mack and I started creating the [CG] world," says Ver non Wilbert, digital set ex tension supervisor. Using measurements from the set, with Wilbert working in Maya, they began at the top of Mt. Crumpit and kept going until they had planned the entire city and its suburbs. The Who-ville site plan shows the town center with streets curving out to form four districts-residential, services, public works, and entertainment-and then extending on out into suburbs that dot the hillsides.

The physical set has 10 buildings. Wilbert and his staff built the rest of Who-ville in Maya-townhouses and condos, a gas station, motel, bank, library, school, theater, bookstore, hospital, a diner, and even suburban tract houses. Thus, when Cindy-Lou's father Lou-Lou Who (actor Bill Irwin), the postmaster general, drives home to the suburbs, you can see other suburbs in the distance; when he looks back, you see Who-ville.
Starting with a terrain map at left, Digital Domain created a storybook world. The snowy mountains and cloud-filled skies in the middle image are all digital, as are the background buildings in Who-ville at right.

To create the world around Who-ville, link it to Mt. Crumpit, and extend that world with the 100 miles of mountain range that Mack wanted, Wilbert started with a NURBS model in Maya. Once the model began to take shape, he switched to polygonal modeling tools, using Maya's Artisan to help create the peaks and valleys for the terrain map, adding extra detail on the slopes of Mt. Crumpit. "We could do high- and low-res frequencies with the polygonal model that we couldn't with the NURBS model," he says. The section of the digital Mt. Crumpit where the Grinch stares "down from his cave with a sour, Grinchy frown" matches the physical set exactly. "We measured the carved foam they used, created a digital version, and placed it in our world," Wilbert says. "If you put a camera there and looked down, you'd see Who-ville."

In fact, Mack did just that to help Howard visualize what he would see once the digital environment was added to the physical set. "When we were shooting on set, we had a low-resolution version of the model and a video tap from their camera plugged into our monitor so every time they lined up a camera angle we could show them where everything would be," says Mack. "We could make adjustments right there on the set."
Using Maya, Houdini, and RenderMan, the crew created this all-digital image. To add a Seussian touch, they placed curlicue clouds created with Amazon Paint in the sky.

The task of turning the environment around Who-ville from a terrain map into a real world made of snow and rocks fell to Johnny Gibson, digital shader lead. "The terrain is part modeled, part shaded, part displaced, and it involves painted texture maps and tweaking until the cows come home," says Matthew Butler, CG supervisor. "[Gibson] came up with a way to procedurally generate a rock-and-snow combination that is mathematically determined at a shader level."

"The terrain is a special animal," says Gibson. "It was developed as a series of multiple [RenderMan] shaders that get merged into one shader." First, using a technique borrowed from lighting, procedures in the shader determined which parts of the surface got snow. "We mark the regions that qualify for snow by the direction the surface face," Gibson says. "We essentially 'shine' snow on the surface as if it were a light." That means if the snow falls straight down onto a surface, more of the surface is covered than if the snow lands on an angled surface. Areas not covered with snow are rendered as rocks.
With a painted sky creating the dramatic mood for this all-digital shot atop Mt. Crumpit, the Grinch and his dog Max can see 100 miles in any direction.

Creating the right amount of rocky detail became a dance between the underlying terrain model and the rock and snow shaders, which used bump maps to add detail for long shots and procedural displacement shaders for rocky surfaces in close-ups. "The shader and topography are so intertwined, it became impossible to separate the tasks," says Butler. The crew worked with Maya (the terrain model), Renderman (shaders), and Houdini; going from Maya through Houdini to Renderman. Within Houdini, they would reduce the polygon count in smooth regions and in other areas add random detail in controlled ways by procedurally displacing geometry. "The model can't have too much detail, but it has to have enough," says Gibson. "If there are too many faces, the snow slides off and it begins to look like a pine cone, but if there aren't enough, it doesn't collect snow properly."
In this shot, the sky is painted. Everything else-rocks, snow, sleigh, Grinch-are 3D. Procedural dynamic systems controlled by animators drive the sleigh down the mountain.

In addition to snow and rock, the terrain is often covered with trees. For shots that called for forests, the effects team considered creating sprites-paintings of trees on simple planes that always face the camera. However, when a camera moves to the side of a sprite card, the card seems to disappear because it's so thin. "We had a lot of shots where the camera was moving so much there was no appropriate angle to use to render the sprite cards, but we didn't want to render a 15 million pine-needle tree," says Butler. "So we created an approximation of a tree."

The tree is made of simple 3D geometry-a trunk with collars encircling it. "Like the collars you see around the necks of dogs that have had surgery," Butler explains. These collars are texture mapped with paintings of branches. To create the illusion of a real tree, Gibson and Charlotte Manning, technical developer, created shaders that use light to suggest more complexity than was actually present. Because light shines through opaque areas in the painting, as the camera sweeps through the forest you can see through a passing tree and into the trees behind. To further optimize rendering, the more distant a tree is from the camera, the less complex is its geometry. "In the opening shot, we have trees from a subpixel size to quarter-screen," says Dan Lemmon, technical development lead, who worked with Manning on RenderMan scripts to do seamless dissolves as the resolution changed.
To give live-action shots an other-worldly atmosphere, the crew added effects such as the stormy sky behind the Grinch looking down from his cave.

But it's one thing to create a pretty background and quite another to have to interact directly with it.

Toward the end of the story, after the Grinch has stolen all the presents, he discovers the true meaning of Christmas just in time to keep the gift-laden sleigh from falling off the top of Mt. Crumpit. He gets underneath and lifts it over his head. For many of the shots in this sequence, the sleigh was completely CG; in wide shots, the entire scene was computer generated. "There's a helicopter-type of shot showing the whole Who valley and Who-ville," says Mack. "We see for about a hundred miles and it's completely 3D."

When the Grinch finally sets the sleigh down after his epiphany, he and Cindy-Lou take the presents back to Who-ville. The terrain in the following 45-shot sleigh ride sequence is nearly always CG, and many of the shots are completely CG. In these, a digital sleigh carrying a big bag of toys careers wildly down the digital mountain with digital Cindy-Lou and Max in the cockpit and a digital Grinch hanging on behind.
In Who-ville, it's always snowing, so the crew added CG snow to shots such as this of the Grinch's dog Max. The background is completely digital.

To animate the sleigh ride, the effects team created procedures that they implemented in Houdini. "The sleigh can't fly down the mountain like a brick. It has to behave dynamically," says Butler. The rule-based system developed by Butler and Markus Kurtz, digital sleigh sequence lead, couples maneuvering behavior patterns, such as banking and pitching that are based on acceleration forces, to position definitions. Using the system, animators can drive the sleigh by setting keyframes to define a path. The system takes that path and projects, in effect, three lines onto the rocky terrain. The middle line positions the sleigh on the terrain, so by comparing points on the left and right lines, the system can determine the bank, pitch, and roll for the Z position of the sleigh. The underlying computer-generated terrain determines the sleigh's acceleration; using math functions with Houdini's Chops editors, the system derives and compares the position of the sleigh from one frame to the next and calculates its speed.

"The big bag on top shifts based on acceleration and position," adds Markus Kurtz. "It leans out as the sleigh goes around a curve, and the smaller objects on the bag are driven by the bag's position." Also, as the sleigh leans, the appropriate ski lifts up. "It's close to real physics," Butler says.

"What I like about the system we developed is that it's not obtrusive," he says. "It's an implicit solution, not an explicit solution. You don't give it parameters and it tells you what the animation is. You give it animation and it tells you where you're messing up."
In Who valley, clouds move much faster than do clouds in real life, and colors are more saturated to help create a real world that's not our own world.

To leave tracks in the snow, displacement maps were projected onto the terrain as the sleigh skimmed along. "During the animation, the snow shader asks, 'have I been collided with a ski?' and if it has, it augments its current displacement," Markus Kurtz says. To do this, the procedural system determines where the collision has occurred and to what degree. In addition, the system drives a particle system so that it looks as if the sleigh kicks up snow as it races down the slope.

When the Grinch is on the top of Mt. Crumpit, and we see for miles in all directions, the landscape is all 3D. However, beyond a certain distance, there is no discernable difference between a 3D model and a painting on a card. "For a big sweeping helicopter shot with the camera moving a lot, you might need 3D for 100 miles," says Mack. "But if the camera is moving 10 feet on crane, you need 3D for only about 100 feet. You can't see a perspective shift beyond that."

With this in mind, the studio used an in-house system called "pan and tile" to help cut rendering time for distant 3D elements. The system determines the cumulative angle of view for an entire shot and then renders one image at very high resolution that encompasses the entire wide-angle view of a background. The studio's compositing software, Nuke, can later create the correct perspective and distort that image appropriately for 3D camera moves-as if the camera is inside a cylinder shooting an image that's been texture mapped onto the inside of the cylinder.
Images courtesy Digital Domain.
With not enough time in a day to put Who makeup on a town full of people, Digital Domain invented a Who Construction Kit to populate Who-ville with digital Whos.

"Traditionally, if you have a 500-frame shot, you have to render each frame, and if the camera move changes, you have to re-render the frames," says Butler. "What we can do is render one still image, and the compositing package will warp it into place for each frame."

This technique allows the studio to create the big, wide-angle shots that help the audience believe that Who-ville is real. To further bolster the illusion, the animators found ways to integrate what they call "atmospherics" into both 3D and live-action shots. For example, 3D volumetric clouds move across painted skies that are used to set moods and punctuate dramatic moments. And in Who-ville, to make the falling CG snow look correct in live-action scenes, the crew modeled CG "hold outs" to match the scene so that the particles would fall correctly in front of and behind people and objects in the scene. Further, the snow, which is rendered with depth of field, sticks when it lands on a Who, then melts. Of course, that effect is a little easier to create when the Whos are also digital-and many of them were.
Throughout the film are CG effects such as this one in which virtual termites crawl through the Grinch's real-but false-teeth.

The amount of time the special effects makeup crew needed each day to change people into Whos limited the number of extras that could be used to around 50 people. However, Who-ville needed a much larger population. The solution was to create digital Whos. "Traditionally, we design and model one character at a time and that wasn't feasible for the 200-plus characters we were adding," says Randall Rosa, character animation supervisor. "So, we came up with a character construction kit."

With the Who Construction Kit, an animator started with a generic man, woman, and child Who model and, using sliders, changed the model's size, age, and facial characteristics to create a new Who. "The Whos have elastic faces," Lemmon says. "Their noses can be anywhere between their eyes and mouth." Noses could also have a twist or a curl; ears could rotate and roll, be flat or stick out.

Once a Who was built, it was dressed in a basic costume picked from a "clothes closet"-hats, shoes, shirts, pants, jackets, and so forth. Martha Mack, creative paint lead, who also painted the skies, worked with costume designer Rita Ryack to create the mix-and-match textures that can be applied to Who costumes. Finally, the Who was given accessories and props. Sometimes the animators started with Whos they or someone else designed; sometimes they created an animation with a stick figure and then applied the animation to a pre-designed Who. Animations created with the stick-figure proxies could be plugged into any Who, fat or skinny; adjustment was automatic. The system could also be used to create "Pee-Wees," 12- to 15-inch characters who live in Who-ville, al though the Pee-Wee seen jumping out of a tuba was created with traditional modeling techniques because he was so close to the camera.
If it seems impossible for the woman in the foreground to hold that big man, it is. Both are digital Whos, as are some of the other Whos in this image.

"This show involves so many different kinds of effects," says Mack. "Not that long ago you wouldn't attempt to do an outdoor scene in broad daylight, in which you see for a hundred miles and it's all 3D. But we also did lots of shots that are one-offs."

For example, digital moths devour digital Christmas stockings and all the CG presents fall out. Santa and his reindeer take off from a Who-ville rooftop in an all-CG scene. The Grinch heads for Who-ville in his rocket-powered sleigh and, when it catches fire, that becomes a CG shot-except for the Grinch, who was rotoscoped out of a practical sleigh and put into the CG sleigh. A CG Christmas tree burns, digital letters fly into a mailbox, CG termites crawl through the Grinch's teeth, and there's a CG tunnel with a real girl and a practical tunnel with a CG dog. In the opening sequence, when Who-ville is first revealed, you can see, mixed in with the actors playing Whos, digital Whos riding digital unicycles, tossing CG presents, and jumping into other digital Whos' arms. The sky is digital, much of the town is digital, and the background is digital.

"We created a magical world that is not on earth; it's in Who-ville," says Butler. "So instead of integrating our shots into live-action shots, it often became the other way around." But when they brought the live action into the digital, magical world, the real-world elements didn't look right. The crew's solution was almost Seussian: "Rather than changing our shots to fit the real world, we'd change the real world to fit our world," Butler says. "It was an interesting irony." And one Dr. Seuss would probably have loved.

Barbara Robertson is Senior Editor, West Coast, for Computer Graphics World.