Animal Magnetism
Issue: Volume: 28 Issue: 12 (December 2005)

Animal Magnetism

When C.S. Lewis's Lucy Pevensie stepped inside an armoire, pushed past the fur coats hanging there, and emerged in the kingdom of Narnia, she uncloaked a fantasy world of good and evil filled with mythological creatures and talking animals-a world that has charmed children for more than 50 years.  In magical Narnia, wolves talk to an evil White Witch, Mr. and Mrs. Beaver befriend Lucy and her siblings, and a lion god named Aslan helps the children rescue the kingdom from the witch's wintry spell. It's a story that seems designed for a cinematic combination of live action and digital characters. And so, it has become.

The Chronicles of Narnia: The Lion, the Witch and the Wardrobe, Walt Disney Pictures and Walden Medias screen adaptation of Lewiss first book in the Narnia series, features realistic CG animals and half-CG creatures acting alongside real people in live-action and digital environments. Directing the film is Andrew Adamson, who honed his skills as co-director on Shrek and Shrek 2, and prior to that as visual effects supervisor for Batman & Robin and Batman Forever.

Tilda Swinton stars as the White Witch Jadis, Liam Neeson as the voice of the lion Aslan, and Rupert Everett as the voice of the fox. Georgie Henley, William Moseley, Skandar Keynes, and Anna Popplewell play the Pevensie siblings Lucy, Peter, Edmund, and Susan. And, a bevy of actors and stunt actors filmed wearing greenscreen tights play centaurs, minotaurs, and fauns.

Dean Wright directed the special effects created by KNB EFX Group and Weta Workshop, as well as the visual effects created at Rhythm & Hues, Sony Pictures Imageworks, and Industrial Light & Magic (ILM), with motion capture handled by Giant Studios. In addition, Wright sent between 400 and 500 shots, primarily matte paintings and composites, to several other studios, including Studio C (in Guatemala), Illusion Arts, Digital Dream, and Sandbox. Rupin Suwannath's crew handled the previsualization.

Aslan, the lion god of Narnia, shown here with actor Skandar Keynes (Edmund Pevensie), was created at Rhythm & Hues. Animators pored through reference footage and photos of real lions to find natural expressions they could apply to phonemes.

Rhythm & Hues was our lead facility and provided all the animation for Aslan, who is CG 97 percent of the time, as well as 90 percent of the battle scenes, which involve hero animation and crowd simulation for a variety of creatures, says Wright. Imageworks got some meaty character animation for Mr. and Mrs. Beaver, the hero faun Mr. Tumnus, the foxes, and the wolves. And they created CG environments. ILM came on later to do a camp scene, the dryads, and a battle sequence.

By the end of postproduction, these three studios were sharing scenes with one another: Imageworks wolf interacted with Rhythm & Hues's Aslan, for example, and ILM's creatures fought on the same battlefield as those created at Rhythm & Hues.

The level of acting our cast exhibits sets the bar for digital effects, says Wright. Andrew [Adamson] is an incredible director; he directed the film for two years, from shooting on location to directing hundreds of animators and voices.

Rhythm & Hues began working on the lion god and battle creatures in 2003 (the same time Weta Workshop began designing the mythological characters), and then began building its digital models in January 2004.

All the characters had a lot of detail, but Aslan had the most involved rig, says Bill Westenhofer, visual effects supervisor at Rhythm & Hues. He was in development for a year and a half. Although an animatronic lion was used during a sacrifice scene that required several shots of the girls hugging him, in all other scenes he is digital.

Before creating Aslan and the cats that would appear in the battle scene, the crew pored through nature videos and books, but it wasn't enough. Nature photographers don't take tight pictures of the corner of a mouth or the corner of an eye, says Westenhofer. So, the studio took a crew to Gentle Jungle, a facility that trains animals for movies. There, a film crew captured a trainer putting a lion through a set of actions that mimicked the tests a rigger would do-a walk cycle, a jump from one object to another, and so forth. Later, the animators would match the camera angle used that day and put their digital lion in motion side by side with the real lion.

Modelers at Rhythm & Hues built the digital lion and other characters with polygons, using the studio's proprietary Voodoo system, but riggers used NURBS tools as deformers and a muscle system with volumes derived from NURBS patches. As with most animation systems, animators working in Voodoo positioned the joints that moved the bones, and the muscles followed. However, to start the muscles moving before the bones moved and, thereby, more closely mimic real anatomy, the riggers added a script that ran after the animators finished.

Rhythm & Hues tangled, clumped, and combed 15 hair types for Aslan's mane, and moved them with dynamics and harmonics.

[The script] evaluates the speed of all the bones and the direction they're moving to know which muscles to fire, explains Will Telford, creature supervisor at Rhythm & Hues. It gives Aslan tiny twitches that help him look realistic.

For Aslan's facial rig, the team combined blend shapes and deformers to utilize the advantages and to minimize the disadvantages of each. Once the animators had created the facial performance, the technical crew used that geometry to drive an anatomically correct muscle system. Then, they drove a duplicate model of the head with those muscles. Lastly, the wizards mixed the two. We combined the animated face and the muscle face so that the original form that the animator intended was there, but it had the muscle movement, Telford says. That extra effort helped create a lion that talked without becoming a cartoon character.

At first, the animators who began working on the lion didn't know who would provide Aslan's voice, so animation director Richie Baneham used Gregory Peck, in To Kill a Mockingbird, as a reference. We needed a character to draw inspiration from, he says. To keep Aslan true to his anatomy even when he talked, the team combed through reference material to find natural expressions the animators could use.

In addition to making Aslan move believably, the crew had to make him look real. Just as riggers and animators compared Aslan's movement to that of an actual lion, the technical animators compared his hair to that of real lions. [Adamson] selected a few lions, and we compared them side by side, says Westenhofer. There are something like 20 million hairs in Aslan's mane and 15 different hair types. That was essential even for a static pose. And, we had to make sure his mane would move in the wind.

Aslan's polygonal body has NURBS-based muscles and a rig with 1400 controls. For his face, Rhythm & Hues combined blend shapes and deformers, and then layered muscle movement on top.

The technical animators ran simulations on guide hairs to produce true dynamics for the mane, but also layered in harmonics to oscillate the hair. By using harmonics, we could play with parameters on the fly, says Telford. Groomers used texture maps to control the hair, and combed Aslan's mane using flow fields and by pulling points on a low-density version of the hair to position it. And, it all seems to work: In an early preview, Newsweek reviewer Jeff Giles praised Aslan as a magnificent bit of computer animation.

After Aslan, of the 40 characters created at Rhythm & Hues, the next most complex was a griffon. He has four legs and two wings, says Baneham. The wings work on a clavicle structure and the legs on a scapula structure that are in direct opposition and compete for the same geography on the body. It was a hard problem. We forced the rig into every situation we could to test it.

For one test, they mimicked a sequence in which a real hawk flew into the scene at the beginning of the big battle. When Adamson saw the test, he switched from a real hawk to the CG character.

Most of the characters in the battle-centaurs, minotaurs, hawks, griffons, ogres, werewolves, cyclops, rhinoceroses, fauns, cheetahs, leopards, and more-marched, flew, scrambled, and fought using motion cycles controlled by Massive Software's Massive crowd-simulation software. Many of the CG characters in the foreground, however, were animated by hand. To make it easier to rig these foreground creatures and for animators to work with them, the character team devised a rigging kit based on 13 snap-together modules for legs, arms, the torso, and so forth.

Motion cycles for such CG animals as the leopards and cheetahs were keyframed. Horses were motion-captured at Giant Studios, as were 20 stunt people ranging from a Cirque du Soleil acrobat to a former football player. It took 22 weeks to capture the needed 6000 motion clips. The centaurs alone had 800 different motions. In all, Rhythm & Hues created 50 Massive agents-25 characters fighting for the good side and 25 for the bad.

Rhythm & Hues created 132 battle shots using Massive, which turned 50 agents into 455,704 characters that acted individually using 6000 motion cycles and rule-based brains. The CG griffon in the foreground was hand animated.

Massive is something that you have to start early, but as soon as you get the motion capture and brain building done, you can blow through shots, says Dan Smiczek, who supervised the process. We created animation for 455,704 characters in 132 shots, all in a short time. There is no way we could have done that traditionally.

The team started by working with a test capture for a faun, to build a basic brain that became a template for general locomotion. Then, a team of 10 people working with the 6000 motion clips used that template to build the rule-based brains for each character. Every character emitted a sound plus information that told the other characters where and what it was: A minotaur had to know when it was fighting a dwarf, for instance.

To integrate the characters into the studios rendering pipeline, they used Massive only for the crowd simulation and rag-doll physics (rigid-body dynamics). We did all the skin binding outside Massive, says Telford. The only thing Massive affected was the joints. The character team simplified the skeletons and rigs, and set up cloth simulations that changed depending on how close the character moved toward the camera. Similarly, the renderer knew when to use extra-low-resolution geometry.

The battle pits good and evil CG animals and mythological creatures against one another. Rhythm & Hues created 90 percent of the battle; Industrial Light & Magic handled the rest.

It was a tremendous undertaking, says Westenhofer. We had bipeds attacking quadrupeds, werewolves running in a loping gait, an animal with six limbs attacking a leopard.

Sony Pictures Imageworks created Mr. and Mrs. Beaver, foxes, wolves, and the half-goat/half-human Mr. Tumnus. The effects wizards also bombed London, created an icy throne room, and melted a frozen waterfall. All the animals talked, and the wolves had to match live-action wolves.

It was all hard, but nothing was brand-new, says Jim Berney, visual effects supervisor. We had to optimize fur rendering to cram these creatures through the pipeline, optimize the shaders, create new combing tools, and match the hair on the wolves, and it was difficult to get the right sense of scale for the ice. Also, the water interaction was tricky.

In addition, the technical crew created an automatic default setting for the fur dynamics that gave the animators a quick idea of how the hair looked and moved. All the animals pelts had several layers. The beavers, for example, had a soft, dense undercoat with coarse fur around their faces and an outer, protective layer with thick follicles. You need the layers so it doesn't look like a stuffed animal, says Dave Smith, digital effects supervisor. The thicker hairs on top catch more specular light, and when the beaver bends toward the camera, you can see the downy layer.

Mr. and Mrs. Beaver, created at Sony Pictures Imageworks, are the most anthropomorphic animals in the film. Imageworks rigged the critters to stand, walk, and swim.

Control hairs determined hair direction, curvature, curl, and shape, but computer interpolation based on parameters controlled thickness and density. Texture maps positioned clumps of hair; parameters defined how dense the clumps were and how tightly they were clumped.

Because the wolves had to match real animals, their fur was the most difficult. A wolf seems darker or lighter, depending on how you're looking at him, says Smith. If you look at his side, you see mostly dark guard hairs. As he walks away, you see into the downy layer, the lighter-colored fur. The foxes needed fewer hair layers, but each hair had more colors, so the crew had to define the color variation.

The faun Mr. Tumnus also needed fur for his goat legs. Actor James McAvoy wore shorts with a fur skirt and greenscreen pants with reflective markers during filming. When his legs were visible, Imageworks replaced them with CG goat legs. We tracked the markers with still cameras on the set by photographing him from a variety of angles and then extracting his position, explains Rob Engle, CG supervisor. That gave us a starting point. Sometimes CG hooves were placed in footprints McAvoy had made in the snow; other times the crew had to create new prints. To sprinkle snow on the faun's fur, the crew used particles made with Side Effects Software's Houdini.

James McAvoy, who played Mr. Tumnus, wore green tights with markers and shorts with a fur skirt so Imageworks could replace his legs with CG goat legs and blend matching CG fur into his skin.

The character team also used Houdini procedures to create chain mail armor for Mr. Beaver. First we did a cloth simulation based on his animation performance, and then we applied a procedure in Houdini that instanced links all over his cloth.

The beavers rigging was the most complicated of all Imageworks animals. They had to move realistically like a beaver, yet stand up, emote, and grab and carry things like a human, explains Smith. Riggers built a beaver body and then had a second setup that allowed bipedal human motion. A muscle system in the characters pushed the skin; however, the animators keyframed secondary motion such as breathing. For the faces, they worked with a fully articulated muscle system that controlled hundreds of muscles grouped to follow the curvature of the face.

The wolves are sometimes real and sometimes digital, so the fur of the CG wolves had to look perfectly real. Imageworks crafted the illusion using dark guard hairs over lighter-colored fur.

These animals are not supposed to be anthropomorphized, says Berney. The wolves and foxes express with their eyes, but they don't use their paws in any way that isn't natural.

The film begins with the bombing of London during World War II, a 30-shot sequence for which Imageworks created a fleet of Heinkel bombers, the Pevensie's neighborhood, flak, and explosions. The shot begins 5000 feet in the air, with the German bombers flying through clouds rendered with Imagework's splat tool for a volumetric effect.

We placed sprites on a particle cloud generated in Houdini, says Stirling Duguid, CG supervisor. The software automatically faces the sprites to the camera. And then we used [Pixar's] RenderMan to generate deep shadows. The sprites were pictures of soft, fuzzy textures.

To create the background, the sky dome, and much of the neighborhood, the painters used Maxon Computer's Cinema 4D for matte paintings; foreground houses were modeled in 3D. For matchmoving live-action elements of pilots in the cockpits, the crew used 2d3's Boujou and Alias's Maya; for fire, they employed both practical pyro elements and Houdini particles.

When Lucy first walks through the wardrobe, Narnia-which is under the White Witch's spell-is wintry cold. Aslan's return begins the thaw. Imageworks created the witch's icy castle, a frozen waterfall that melts, and a river that begins to flow.

We first see the castle from a distance when three of the children watch their brother Edmund follow the witch. (The witch had lured Edmund to the dark side with Turkish Delight candies.) In the foreground, the children, filmed on greenscreen, became one layer; the next layer was 3D trees, and then beyond the trees were the castle, mountains, and hills, created with simple geometry painted in Cinema 4D.

Everything except Edmund is digital in this wintry environment, one of several created at Imageworks. The studio also extended an ice palace set and handled shots with a frozen waterfall and a melting river.

Inside the castle, Imageworks extended a set, built with backlit plastic, using 3D models rendered in RenderMan with subsurface scattering, ray marching, raytracing, refractions, light diffusion, specular light, and light glows to match the fake ice. Working with the ice required subtlety. It's a challenge to go from ice that feels like the scale of an iceberg outside to the throne room, which is closer to an ice sculpture, says CG supervisor Mark Lambert.

The frozen waterfall sequence, which involved characters interacting with ice and water, was even more complicated. For this, the crew created the water surface with Houdini particle simulations brought into RenderMan and composited layer by layer in Imageworks Bonsai.

Each effect is its own layer, says Duguid, water drops, snow on fur, footsteps on the ground, the ice. The waterfall is sometimes a quarter-scale miniature and sometimes a 12-scale miniature created at New Deal. The water ran over the quarter-scale model, while the children hung onto a practical ice chunk in a water tank as a wave of water crashed down upon them. Imageworks extended the wave and created shots underwater, with the Beavers and stunt-double children on the ice chunk moving downstream.

We've done some pretty hard stuff, says Berney. One reason it's hard is because we came a little late to the party and didn't have preproduction time to create the characters. But we didn't come as late as ILM.

When the White Witch makes a deal with Aslan to save Edmund's life, crowds of goat-legged creatures, created and animated at ILM, surround them. Later, when the fighting starts, ILM animated a variety of creatures in hand-to-hand combat. In addition to the crowd scenes, technical animators using particle systems created dryads-women formed from cherry blossoms, willow leaves, and apple blossoms.

This is the most labor-intensive show I've been involved with, says Scott Farrar, visual effects supervisor at ILM. There are dozens of characters in the frame, and they all have to look good and have ground shadows. I've never had a show like this where there was so much detail in any shot.

ILM had less than 12 weeks to create and animate 30 creatures with five variations for each-150 in all. The characters all required cloth or fur simulation, or both.

There are probably hundreds of little hand-to-hand combats, points out Jenn Emberly, animation director. We provided a backdrop to the fight between Peter and the White Witch. [Adamson] wanted to achieve a sense of chaos in the background while maintaining the focus on the fight. It was about movement through the screen. The animators started by deciding which creature would fight which creature, what actions the characters would perform, and whether they would move across the screen or fight in one place. Then, the animators choreographed what Emberly calls a violent dance: layer by layer, step by step, adding more and more detail as the shot progressed.

We used everything we could, says Emberly. We tried to create a library of motion cycles to cover our bases, and then have variations so that everything would look lifelike with lots of variety.

To make it easier for the 20 animators, the riggers tried to apply the same set of controllers to all the creatures. The animators performed the primary animation using a rig that had muscles enveloped to a skeleton. Then, an additional muscle system could add muscle flexing with specific timing.

With so much detail in the shots, Emberly found it hard to stop finessing. We had to go through the sequences and prioritize what we would trick out with the muscle system using the simulation, and when we would add shapes on top, says Emberly. We didn't have time to do that with all of them.

To help keep their creatures consistent with those developed at Rhythm & Hues, ILM used the same motion-captured files, and the two studios exchanged QuickTime movies of the animation. It's been nice to share that kind of information on an artistic level, says Emberly.

Has all the work been worth it? Farrar believes it has. It's a big step to surround live actors with CG critters, he says. And it's the first film to go beyond Babe, where you had replacement mouths. Financially and artistically, it's a big jump, but the story lends itself to that perfectly. You have CG animals that offer the audience what I always want: great characters with a great story. It's an interesting film to watch. It's something you haven't seen before.

If Farrars correct, it's only the beginning of the adventures in Narnia.

Barbara Robertson is an award-winning journalist and a contributing editor for Computer Graphics World. She can be reached at