Imageworks new tools and technigues to replicate 2D graphic style in a 3D world.
Two improbable buddies—Boog, a 900-pound grizzly bear voiced by Martin Lawrence, and Elliot, a one-antlered mule deer voiced by Ashton Kutcher—exhibit the strong graphic shapes with lush textural details used in Open Season, an animated fi lm created with 3D tools.
If a summer filled with CGI features has taught us anything, it’s that moviemakers can’t guarantee box-office success simply by using 3D to make an animated film. The thrill of seeing 3D characters in a 3D world is gone, which might explain why CGI features are beginning to look so different from one another. Or, maybe it’s rampant creativity. Other than the fundamental rule “You must start with a good story,” there is no CG aesthetic, no recipe for what makes a good CGI film. But if there were, Sony Pictures Animation’s (SPA) first foray into the growing field of CGI animation would have changed it. Directed by Roger Allers, Jill Culton, and Anthony Stacchi, Sony Pictures’ Open Season is layered with improbabilities, artistic and technical.
Set in Timberline, a Northern Exposure type of backwoods town, the film stars Boog, a 900-pound grizzly bear voiced by Martin Lawrence. Forest ranger Beth (Debra Messing) raised Boog from a cub. He lives in her garage and performs in her nature shows. At night, Beth tucks Boog into bed with his little teddy bears.
One day, a hunter named Shaw (Gary Sinise) rolls into town with a one-antlered mule deer strapped to the hood of his truck. Remarkably, the deer, Elliot (Ashton Kutcher), is still alive. He persuades Boog to free him, and follows Boog home; and that’s the end of Boog’s cushy life. The fast-talking Elliot tempts Boog into trouble, and the rangers send the two miscreants into exile in the woods. There, the citified bear and hapless deer discover true friendship while encountering wacky woodland critters. But, it’s hunting season. When Boog realizes he’s among the hunted, he helps turn the tables on the hunters. And then, he must decide whether to return with Beth to Timberline or live as an animal in the woods.
“It’s a good, universal theme,” says Culton. “Kids have to leave home, whether for kindergarten or college, and parents have to watch their kids grow up and be OK with letting them go.”
A Fungible Pipeline
Making its first animated feature was a learning experience for Sony Pictures Imageworks, as well. SPA relied on Imageworks’ visual effects pipeline to produce the animation, and many of the artists and technical crew at Imageworks moved between visual effects shots for clients and work on the feature animation. But, Open Season bears little resemblance to Imageworks’ Oscar-winning visual effects in Spider-Man 2, its Oscar-winning short animation, “The Chubb Chubbs,” or the animated films The Polar Express and Monster House, which used a motion capture-based pipeline to drive much of the human characters’ performances.
“We wanted to go for more of a graphic style,” says Culton. “Our characters have strong shapes with lush details; the backgrounds are the same way. It’s a challenge in 3D to find a graphic look that worked from different angles. We had to come up with special tools for the backgrounds and animation, and sometimes we had to cheat.”
Culton discovered the graphic look that became the film’s style in a book stored in her attic: The Complete Christmas Card Art of Eyvind Earle. During the 1950s, Earle had painted backgrounds for such Walt Disney animations as Peter Pan, Lady and the Tramp, and Toot Whistle Plunk and Boom, and was a color stylist for Sleeping Beauty. By pushing the background out of focus, using long, raking shadows, and a simple graphic style that captured only the essence of environments, Earle put the viewer’s attention on the characters.
Thus, Imageworks not only needed to add tools such as those for layout artists to its pipeline. It also had to modify some of the existing tools and techniques to replicate in 3D the strong 2D graphics that defined the visual style for the film, both in the backgrounds and in the characters. “We leveraged off Imageworks’ expertise, but this film’s needs were quite different,” says Doug Ikeler, visual effects supervisor for Open Season, who led a team that grew to 250 people during production. A look development package, for example, helped artists quickly shade, light, and render the numerous ordinary elements that added details to the 3D environments.
“In visual effects, we have texture artists, shader artists, and lighting artists working on elements, and that’s a slow, iterative process,” Ikeler explains. “We needed a system that one artist could use to put a book in the corner of the room in one day. We used it mostly for props and one-off stuff, not for main characters, trees, or anything with heavy re-use.” With the new tools, the look development team created approximately 600 digital props.
But that change to the system was minor compared to the major effort and attention the crew focused on converting Eyvind Earle’s graphic style, as interpreted by SPA’s art department, into 3D backgrounds and characters.
Behind Every Character
The first sequence to move into production at Imageworks was a scene in which Boog flies to the woods in a helicopter. When Ikeler had first spotted it, he noted that because of the long forest vistas, the backgrounds would probably be matte paintings with a little 3D geometry in the foreground.
“Everyone was looking for footage; because the layouts were simple to do, they put this sequence into production first,” Ikeler says. “It wasn’t a good idea.” The problem was the painted backgrounds defined the look for the sequence. “When we put grass in the foreground, everyone recognized that the 3D foreground and the 2D background didn’t fit together.”
Something had to give. The 3D team decided to work on one tree from the artwork, a Douglas fir, rather than the entire forest, to work out a way to add detail without losing the graphic style. The pine tree looked like a Christmas tree, with swooping branches, pinecones on the outside, and needles on the inside.
“It’s easy to paint a pine needle texture on a tree that has a smooth profile on one frame, or to render one frame of a 3D tree, but the challenge was to keep the texture and a smooth profile when the camera spins,” Ikeler says. “It was a huge R&D project.” The system Imageworks developed used layers of cards fashioned into the shape of a tree, each with clumps of pine needle textures. As the camera moved around, the cards also moved so they laid flat against the underlying structure.
Once everyone approved that tree, the crew created other tree types and ages, varying the levels of detail on the cards for each. Trees close to the camera used cards with more detail than those that needed to blend into matte paintings in the distance.
The forest floor presented a different challenge. “The directors expected to move or add ground cover on any shot,” says Ikeler. “They wanted control through the final composition.”
Five types of ground cover—pine needles, dirt, clover, dead grass, and green grass—defined with painted maps provided the foundation for landscape artists who could then add 3D rocks, tree roots, mushrooms, leaves, and 10 different flower types with a new General Instancing Tool (GIT). The GIT artists dotted the ground with particles that grew into 3D elements or sprites, which were painted cards.
Imageworks art-directed water shots using a simulation hierarchy. First, a coarse simulation directed water flow (at top, left). Then, smaller simulations created foam (at top, right). Custom Houdinibased plug-ins handled interactions procedurally (bottom).
Even the water was highly art-directed. “Water is always a hard effects element,” says Ikeler. “We needed all the cool bells and whistles—splashes, foam, water running down things, and character interaction—but we knew it had to be art-driven.” In the most dramatic shot, Boog breaks a beaver dam. The resulting flood wipes out the other animals’ homes and flushes them into the hunters’ grounds. The sequence, which they’ve dubbed Splash Mountain, sends the animals down rapids while they cling onto logs.
“One of the major issues was that we had preplanned marks, cues, and beats to hit,” says Dave Stephens, effects animation supervisor. “In a lot of cases, you set up a simulation, run it, and work the shot around it. This was the reverse. They had the shot already worked out; we had to work to those requirements.”
The effects team devised a Side Effects Houdini-based system to handle Splash Mountain, and from that, derived water for the calmer streams and ponds. The solution used a hierarchy of simulations. “We moved a mass of water through the scene using a coarse-level simulation,” Stephens explains. “Then we fed that data into smaller scale simulations. At the end of each simulation, we had a set of particles, and we could construct a surface or have motion vectors that described the flow. The large simulation directed the overall motion; the smaller simulations directed small areas and created foam elements.”
Also, a system of custom plug-ins for Houdini controlled the interaction of the water with static and moving elements. “The system monitored how much water flowed past an intersection and decided whether there should be foam or not,” says Stephens. If a large splash was called for, 3D foam flew up into the air as particles; otherwise, 2D particles traveled on the surface to create foam that flowed with the water. The system could also tell whether a character or object was entering or exiting the water and create appropriate deformations, splashes, and ripples. “Because this part of the system was entirely procedural in Houdini, rather than a fluid simulation, if a director wanted something physics couldn’t dictate, we could make that happen,” he adds.
Boog, Elliot, Beth, and Shaw head a large cast of secondary characters—a herd of deer, beavers, rabbits, a skunk, a squirrel, and other animals, plus a pack of hunters and an old sheriff. The Eyvind Earle look chosen for the backgrounds influenced the design and animation of these characters, as well. All have exaggerated shapes: Boog has a big belly and small head, Elliott has a large head on a long neck, the beaver has a big nose. All the men except Shaw are round; the women have elongated faces.
Similarly, the town of Timberline and the vehicles the human characters drive are highly stylized. “All the buildings are a little bit off-center and there are no parallel lines,” says Ikeler. “At first it was tough getting the modelers not to build everything perfectly.”
Other than Main Street and Beth’s garage, most of the action in town takes place inside cars and trucks. “Making the cars fit in the wonky town was a challenge,” says Max Bruce, CG supervisor, “but one of the ways we tried to evoke a 2D feel was to give vehicles a mid-20th century look.” The hero vehicles, however, were more up-to-date, albeit simple and iconic.
For the characters, the design challenge was retaining their simple lines while providing enough textural detail to make them interesting. For the humans, that meant putting enough detail in texture maps so that the humans looked real, but not so much that they looked realistic. “We couldn’t put real skin on a 2D character,” Ikeler says. “But we needed to please the viewers and give people what they’re used to.”
For animals, it meant art-directing fur on a cast of furred characters. At first, in keeping with the graphic look, SPA didn’t want the animals to have realistic fur. That plan changed when the look development team showed the directors the fully-furred characters. “When you see Boog’s fur, you want to get in and hug him,” says Allers. “And you can see the growth patterns, the swirls where the direction changes in the middle of his chest.”
Imageworks uses a hair and fur system based on control hairs interpolated during rendering into millions of hairs and texture maps that control how the hair looks. The grooming tools are proprietary; the studio uses Autodesk Maya dynamics for hair simulation.
For Open Season, CG supervisor Darren Lurie helped create new tools that could clump and mat the hair in specific designs to achieved the dramatic graphic look required for the animals and the humans. “We could create geometry and fill the shape with hair,” he says, “which was a good way to visualize the overall shape we were trying to match in the drawings.” By designating areas of influence around a control hair, combing artists could cause nearby hairs to clump around it. Each of the clumps could have different properties and dynamics.
New tools helped groomers comb the characters’ fur into specific styles; each clump could have different dynamics. For further art direction, a volume-based rig could also move the fur.
Once groomers finished combing the hair, the hair-simulation team incorporated the animation, and the hair moved dynamically and automatically as the character moved. However, for some characters in this film, because the movement of the hair affected the silhouette and, therefore, the graphic style, the crew needed a way to direct the movement. “We needed more control than we previously had,” says Chris Yee, lead technical director, who had worked on hair for The Polar Express and fur for both Stuart Little films. The answer was a volume-based rig.
The rig was the brainchild of Dustin Wicke, Francois Chardavoine, and Armin Bruderlin. “The volumes, which were part of the rig, moved the hair as a transform when the skeleton moved,” Yee explains. “There were no dynamics happening at all.” Beth, for example, had separate volumes to control locks of hair on her face and for her ponytail. The volumes were bound to a joint system that was bound to the skeleton. Each volume drove a specified number of control hairs that, in turn, drove more hairs, and eventually generated thousands of hairs on the surface.
“The volumes were easier to manipulate than trying to grab each hair,” says Yee. “Some of the hairs have 50 CVs.” For characters that used these volumes, Maya’s hair-simulation engine added ancillary movement later, but the volumes always controlled the goal curves.
The two-dimensional backgrounds and character designs also influenced the animation style and, therefore, the character rigging. “We had to treat the animation in a way that would match the style of that world,” says Sean Mullen, animation supervisor. “To do that, we took more of a 2D approach by using shapers on the rigs.”
Usually, animators pose 3D characters by rotating joints. However, the animators working on Open Season could also move control points on the surfaces of characters to, in effect, sculpt poses. “Imagine that an arm was divided into eight slices that looked like little squares,” Mullen says. “You could grab a square and move that section of the arm like a rubber hose. Or, you could grab points on the square to pull the surface on one side.” This pliability enabled the animators to mimic the squash and stretch of traditional animation. Working in 3D, they could stretch a character’s arm to exaggerate a gesture, and then snap the arm back into its original shape.
“Early CG characters often felt ‘puppety’ because the animators didn’t have flexibility,” says Allers. “The new shaping tools not only helped us achieve a graphic look, but they gave us the strength of traditional animation: greater stretch and squash and strong shapes.”
Giving animators the flexibility in 3D that their counterparts have in a 2D world brought with it the same downside: It became easy for them to go “off model.” When that happened, Mullen used new drawing tools to correct shapes and tweak timing. “I could draw over the renders,” he says. “Sometimes I’d go over an entire character, to show how to pose it, what kind of timing to use, the arcs to follow, the shapes. It was basically like doing traditional animation.” The animators could then use the shapers to match his drawings.
Light and Dark
While creating a performance, the animators also needed to be aware of lighting in the scene. In Eyvind Earle’s illustrations, background colors and long, raking shadows help focus attention on the characters; similarly, in Open Season, the graphic shapes of shadows often had as important a role as solid objects. “We used lighting thumbnails, the little paintings that showed the lighting and colors in the scenes, as reference,” says Mullen. “We’d know, for example, that if we put Boog all the way to screen left in a shot, he’d go completely dark.” If a gesture on an approved animation, however, moved into a shadow, the lighting team would add highlights.
Colors also affected performance. For example, in bright areas, little Elliot could be in front of big Boog because his lighter colored fur created an internal silhouette, but Elliot’s color values were lost in the shadows.
Because shadows played such an important role in the look of the film, the crew developed techniques to add shadows that were not cast by geometry and to exaggerate or even remove the cast shadows. If this had been a typical 3D animated film, to light Beth’s garage, for example, the technical directors might aim lights based on where the characters were standing to create a particular look, and then render the shot from various camera angles. But because this film was so art-directed, the lighting crew had to design shadows not from the geometry, but from the effect they had when seen in the 2D plane from the camera.
Animators could move control points on the surfaces of characters to sculpt poses and mimic traditional animation’s squash-and-stretch style.
“At first, we tried to push the virtual set idea onto the directors,” says Ikeler, “that you’d build and light the set once. But all the notes we got from the directors were image-based. They wanted small edits that kept taking us back to a 2D art place.” For example, the directors might ask for a gradient on a tree in the top right corner of a frame, or less light on the chevron on Boog’s chest. “For us to make those changes by moving the lights in 3D meant huge render times,” he says. Again, something had to give.
The crew rendered every object for every type of light, creating separate lighting passes for fill lights, key lights, ambient light, shadows and so forth. They used RenderMan’s AOV (arbitrary output variables) option for the multipass rendering. “When we put all those separate objects together in the composite, we’d get the same result as if we rendered them all at once,” Ikeler says, “but everything, for the most part, was still tweakable. It allowed us to do 10 iterations in less than a week and address tiny notes on the shots every day.”
Effects artists created fi re for this explosion using Maya’s fluid simulator, tweaked it in Houdini, and rendered it in RenderMan with a volumetric shader.
In addition to tweaking the lights, the artists could also change individual elements in a scene during compositing. If the directors didn’t like having both of Elliot’s legs the same brightness, for example, a soft gradient could darken a rotomatted leg.
“We used a lot of grads,” adds Bruce. “We did vignetting to make the trees and background darker toward the bottom, and to push anything that approached the perimeter of the screen darker and out of focus.” Also, depth passes adjusted during compositing helped reduce the contrast, darken, and decrease the focus of objects in the distance. “We could select parts of an image and do whatever we wanted,” Bruce says. “It helped the 2D look.”
You might ask why SPA didn’t ask Imageworks to create the film in 2D rather than 3D if the goal was to achieve a 2D look. The answer is simple. It isn’t just that the studio had a 3D pipeline in place to work with, it’s that the film wouldn’t have looked the same. Boog wouldn’t have been furry and huggable, and Splash Mountain wouldn’t have become a wild, three-dimensional thrill ride. And, as is evident by the increasing numbers of CGI features, audiences now expect animated characters to be 3D and live in a 3D world. By finding ways to incorporate 3D graphics into a 2D graphic-shape language, Imageworks blended the best of both worlds and created a unique CGI feature with a strong visual style.
Barbara Robertson is an award-winning writer and a contributing editor for
Computer Graphics World. She can be reached at BarbaraRR@comcast.net.