Animal Kingdom

More than a thousand Narnian years have passed since the Pevensie children walked through a wardrobe and into the fantastic kingdom of Narnia, and yet, in the storybook's real world, Peter, Susan, Edmund, and Lucy are only a year older. Aware that the actors playing the children in The Chronicles of Narnia: The Lion, the Witch and the Wardrobe would soon grow up, filmmakers moved quickly to re-create the kingdom with all its digital mythological creatures and talking animals.

Visual effects supervisor Dean Wright, who received an Oscar nomination in 2006 for the first film, began working with returning director Andrew Adamson that same year on Disney's sequel:  The Chronicles of Narnia: Prince Caspian. This time, knowing the film would have 1600 visual effects shots, Wright split the work with co-visual effects supervisor Wendy Rogers who, most recently supervised visual effects for the animated feature Flushed Away and, before that, had worked with Adamson on Shrek.

"The work on this film is markedly more complex than the last film," Wright says. "We have more characters, a greater variety of characters, and the interaction between humans and CG characters has been amped up."

After 1300 years, only two of the previous film's main characters still exist in Narnia: Aslan the magnificent lion who created Narnia, and the evil White Witch played again by Tilda Swinton. This generation of Narnians-the Minotaurs, centaurs, satyrs, and various animals-resemble those from before, but they're grittier, wilder looking. As the story opens, evil uncle King Miraz and his army of Telmarines hold the rightful ruler of Narnia, Prince Caspian, under siege. But, help is on the way. Using Susan Pevensie's magic horn, Caspian summons the children, who return to Narnia and help the Old Narnians reclaim the kingdom. 

As the story progresses, the children meet four main Narnian characters: a large swashbuckling mouse named Reepicheep (Eddie Izzard), Bulgy Bear (David Walliams), a badger named Trufflehunter (Ken Stott), and, of course, Aslan (Liam Neeson).

"This show represents the most shots MPC has ever taken on and certainly the highest complexity," says Greg Butler, visual effects supervisor. "We had furry creatures, exacting performances, crowd work, environment work...a bit of everything."

The Narnians appear in the foreground, midground, and in armies in the background, each built to jump into a hero role if needed, with dynamic fur and a facial performance. For the real animals, deer, foxes, wolves, cheetah, tiger, boar, bullmastiff, German shepherd, rabbit, and other species, the crew referenced photographs and video­tapes, building and rigging the models in Autodesk's Maya, using Pixologic's ZBrush for displacements. Although MPC received the 3D assets from the first film, the Narnians had changed enough and the studio's technical approach varied enough that modelers rebuilt the creatures from scratch with two exceptions: Modelers used the griffin and Minotaur leg models built by Rhythm & Hues as a starting point for Caspian. 

A standardized rigging approach helped speed the character setup process. "We can build a quadruped or biped pretty much in a day as a starting point," says Butler. "We check it out in animation, and if there are no changes, animators can be working on it two days later." For muscle dynamics, the studio used Autodesk's cMuscleSystem, a Maya plug-in developed by Comet Digital. Custom deformers jiggled the skin.

By contrast, the process of creating mythical Narnians began on set. The crew filmed actors wearing tracking markers on green stockings, with witness cameras-two cameras on either side of the film camera-to give rotoscopers and leg animators side views. Then, the animators performed CG models of goat legs for fauns and satyrs, a bull body and legs for Minotaurs, and half a horse for the centaurs that compositors spliced onto the actors.

In the previous film, to raise the stunt actors playing centaurs horse-head high, the crew put them on platforms covered in green, but that limited their range of motion. This time, the actors wore Powerisers-springy curved metal devices that gave them height and mobility-on each foot. "It just helped make the world feel a little more vibrant," says Rogers.

Sometimes, a little too vibrant. "The Powerisers gave an element of the ups and downs of horses walking, but sometimes the actors were too zealous," says Greg Fisher, who, along with Adam Valdez, co-supervised a crew of 75 keyframe animators and 18 crowd animators. "We needed to make a centaur feel like one creature, not a horse with a human stuck on its neck. If the live-action actor moved around too much, it was difficult to get the blend from the human to the CG horse. An actor can do a sharp turn on the spot, but the back end of the horse had to spin around 180 degrees."

For the Telmarine army and background Narnian satyrs and Minotaurs, animators worked from human motion-capture data adapted and reconfigured, in the latter cases, to fit onto rigs for goat legs and four-legged quadrupeds. Giant Studios did the motion capture for MPC and for Framestore CFC's Telmarines, including capturing horses and, for MPC's background centaurs, stunt actors strapped onto a horse's neck.

At one point during the design process, the 22-inch-tall Reepicheep began looking like a tall, thin rat. A combination of sculpting, skinning, a pose guide for animators that gave him a little bend in his back, a belly, and an apron of fur that hangs between his legs turned the creature into a more endearing, if rather large, swashbuckling mouse who appears in approximately 100 shots. "We also took liberties with limb length and position," says Butler. "He had to sword fight. And, to keep him from looking too cartoony, we kept the movement of his face fairly far forward."

The challenge for the Reepicheep animators was in tuning a performance that could run from aggressive to mocking to mild. "Trying to retain the mousey qualities even though some of his lines are quite aggressive was the fun bit," says Fisher.

Bulgy Bear, on the other hand, was consistent in his timidity. "He's always trying to find his inner bear," Fisher says. "He tends to stick his hand in his mouth to cover his nervousness, but as the movie progresses, he starts to stand up for himself."

For the mouse hair, Bulgy Bear's coat, and for fur on the other Narnian creatures, MPC used its proprietary Furtility program, which they had revised for 10,000 BC and revamped further for Caspian. To make the fur program artist-friendly, the studio had developed a texture-map based grooming system using Maya's paint tools rather than guide hairs. "The interactivity was really good," Butler says, "especially for small and medium-size fur."

Although Butler initially thought the trees would be most difficult to render, fur rendering proved harder. "We had to light different types and styles of furry creatures, including versions of real things on set in all kinds of lighting conditions," he says. On set, the production captured HDRI maps using Hoyt Yeatman's HDR camera, which captures bracketed images simultaneously from four directions. "The results were great," Butler says. "We could use those maps for reflection maps, and we could also run them through an algorithm that produced CG lights with the same colors, intensity, and distribution as the HDR map described on set. We could ask for 64 lights or any power of two, and it generally gave us good ambient lights. Then we added CG key and rim lights for fine control." 

The biggest problem for the studio to solve, however, was the volume of work. "It wasn't one specific effect or part of the show that was hard," Butler says. "It was the sheer amount and variety. But, we had a plan and saw it through in a fairly short production schedule."

Framestore CFC: Aslan, Trufflehunter
Two of the main dialog characters, Aslan and Trufflehunter, landed at Framestore CFC where Jon Thum supervised a visual effects crew of approximately 120 who created that studio's 550 shots. Each of the characters had close to 150 shots, and each had particular demands. For Aslan, the studio needed to achieve the same level of sophistication as had Rhythm & Hues in the previous film, and more interaction with the actors. Trufflehunter had important acting scenes and interaction with environmental elements.

"Lucy and Aslan interact physically," says Rogers. "They hug each other and roll around together, and then they have a subtle emotional dialog. In some cases, Aslan has to emote in close-ups without speaking. The work in the first film was quite stunning; they set the bar high. But Framestore did an excellent job to meet and, in some cases, exceed that work."

"We use a network-based system with a variety of filters that perform small operations on the fur," says Mike Mulholland, CG supervisor. "Each filter can accept texture maps that control length, width, scraggle, bends, noise on the fur, and how clumps interact."

Even so, creating interaction with the actors was difficult. For the scene with Lucy, a crew member wearing a lion's head played the role of Aslan. "She pushes his fur around and puts her arm through the mane," says Thum. "We used simple shapes to push our fur out of the way rather than trying complex actions because you go crazy when you try that. Unfortunately, our CG lion couldn't completely occlude the guy with the lion's head, so we had to paint in live-action areas of her clothes and some of her face."

The studio started on the shot early, knowing how difficult it would be to track Lucy, track and remove the man wearing the giant head, animate the lion with the proper weight and timing for the tumbling scenes, cause the hair to react properly, and paint the missing parts. "These shots were the first ones turned over to us because the interaction was so important," Mulholland says. "And they were some of the last ones we delivered."

As did MPC, Framestore CFC used the HDRI images from Hoyt Yeatman's camera. "We got four images at different exposures, proper 360-degree images, stitched and calibrated," Mulholland says.

For texture and shader reference, the crew turned to the first film and footage of Zion, the real lion that had been Rhythm & Hues' model, to build Pixar RenderMan shaders for Aslan's fur. "To handle the long fur, we calculated our own fur occlusion per frame," Mulholland says. "It's like ambient occlusion, but we are doing that hair-to-hair to get shadowing deeper within the fur. Because we calculated it per hair, as the hair parted and moved, the occlusion changed." The process is quick-Mulholland estimates it takes between 10 and 15 minutes for the calculations, which they then fed into RenderMan. 

For rigging, Framestore CFC also used the cMuscleSystem on the body, but their own muscle system, FCmuscle, for Aslan's face. "We did have shapes on top," says Mulholland. "You need both mixed together, but the animators mostly drove muscle groups rather than shapes." A shared GUI between Aslan and Trufflehunter helped animators move easily from one to the other.

Trufflehunter's most demanding action scene found the three-foot-tall badger running through two-foot-tall ferns. "We had tons of interaction with plants in the plate," says Thum, "plus we had to replace the plants with CG ferns." In addition, during the sequence, Prince Caspian slings the badger onto his shoulder.

On set, the actor carried a badger-shaped black bag that the studio replaced with the CG badger. "We pretty much designed the badger ourselves, so that was quite tricky, but once he was made, he was easier than Aslan," says Thum. "One thing that was difficult, but that we solved early on, was how to get his eyes to read." The solution was replacing the black fur around his eyes with a grayish ring animated in compositing when Trufflehunter is shadowed.

In addition to Aslan and Trufflehunter, Framestore CFC created dryads made from petals with Side Effects Houdini particles, unwrapped a magic tree, which provided a portal from Narnia to the real world, built a London Underground station that transitions procedurally with Maya scripts from the real world into Narnia, and used Massive software to send an army of Telmarines into a river. "We ended up doing the interactions with the river surface using 2D elements," says Thum. "It was painstaking to put all those elements in, but it worked quite well."

Weta: Bear, Werewolf, Ice, Backgrounds
With Weta Digital's award-winning The Lord of the Rings and King Kong experience in blending live-action and CG characters into miniatures, the studio was an obvious choice for the big sequences during which the Narnians try to sneak into King Miraz's castle. The studio also encased the White Witch in ice and handled stand-alone shots for which they created a bear and a werewolf.

For the characters, Weta used a new companywide facial system that creates target shapes that preserve volumes. "For Gollum, we did that with artistry and careful consideration," says Guy Williams, visual effects supervisor. "Now, target shapes slide over bones and muscles, and the volume preservation happens automatically. All the animator does is tweak the final shape."

The system creates the animation blendshapes after riggers line up muscle end points based on a human or canine face (so far) with a specific creature's geometry. The volume preservation takes place after animators working in real time within Maya create the facial performance. "This facial system gives us so much more control," Williams says. "We can spend more time on performances than on the mechanical aspects."

That was especially important for a werewolf that pokes his nose out of a hooded cloak and delivers a sinister speech.

Weta's bear is not a typical Narnian; it threatens little Lucy. "It's an angry bear looking for food," says Paul Story, animation supervisor. "We had to exaggerate the animation to sell the weight, but other than that, he was straightforward."

The werewolf, however, was not so straightforward. On set, an actor wearing a cloak delivered the werewolf's speech, so Weta needed only to fit the CG wolf's muzzle inside the hood; but during the sequence, the werewolf throws off his cloak and fights with Edmund. Production started with an actor wearing a prosthetic upper body and head, and green pants to enable CG leg replacements. "As we went forward, we wanted longer arms, so we replaced those," says Williams, "and then because the CG face could do more than the servo, we replaced that. We tried to save the cloak, but it compromised the performance." For the fight, an actor in a blue suit provided the animators with reference motion for interaction with Edmund.

During that sequence, the White Witch appears inside an ice wall. To create the ice, Weta rendered 20 passes in RenderMan that allowed compositors working in Apple's Shake to control the diffusion. By setting various ray depths for the passes, they built the volume more efficiently than by raytracing an entire volume all at once. "We do the math in the comp," Williams says. "The result is still accurate, but we don't have to fire 64 rays to get diffusion through the depth."

Tilda Swinton's image, augmented with CG to give her an ethereal look and then composited, floats inside the icy wall. But she reaches through the wall for Caspian. "We basically rotoscoped Swinton's hand off the plates and then treated it with 3D ice elements to connect it to the wall," explains Colin Always, compositing supervisor.

Some of the studio's slickest work, however, was in blending miniatures shot in various scales into a seamless background. "We had a few shots with a 10th-scale miniature in the foreground attached to a 24-scale set with a 100-scale miniature on the other side," says Williams. The 10th scale was of castle towers, the 24-scale castle was 20 feet tall, and the 100-scale landscape was the size of a basketball court. Once Weta received plates with shots of these miniatures, it tracked the cameras using Science-D-Visions' 3D Equalizer, pulled that information into Shake, and used the camera moves to drive two-dimensional transformations.

"We used standard 2D transforms as well as corner pins and warps," says Dan Lemmon, digital FX supervisor. "Essentially, we compute the difference between two plates and try to figure out, using these tools, how to slot the plates together. We also used Maya to slot various pieces together manually." In addition, Weta augmented the backgrounds with set extensions, and composited MPC's Narnians into the shots.

Meet Me at the River
To create a river guard, the production brought in Scanline fairly late during the process. "We were impressed with their simulation tools," says Rogers. "The river guard is one of those amazing effects where you try to wrangle a natural element to do your bidding. It's a force of nature temporarily formed into a vaguely anthropomorphic shape that responds to Aslan's bidding."