The setting of the $250 million The Golden Compass, based on a novel by Philip Pullman, is an alternate world where humans’ spirits are embodied in animals – talking “daemons” that follow them everywhere – and polar bears are ferocious armored creatures with a kingdom of their own in the arctic north.

For the sake of the book’s legion of fans, as well as for the general audience unfamiliar with the source material, it was crucial that the film’s multitude of animal-like characters be executed with photorealistic panache, and that the many CG environments appear seamless on screen. A slew of VFX houses in the U.S. and the U.K. were brought in to work with the film’s overall VFX supervisor, Michael Fink, on various aspects of the film’s critters and its settings in a futuristic Victorian London and elsewhere.

F&V spoke with the three lead facilities on the project – London’s Cinesite and Framestore CFC, and Marina Del Rey, CA’s Rhythm & Hues – to learn what was involved in bringing Pullman’s otherworldly visions to life on screen.

The most obvious challenge, because it necessarily affects almost every shot that also features human characters, was the creation of the daemons. Daemon duties were divided between Rhythm & Hues, which was tasked with creating the primary daemons for the film’s lead actors, and Cinesite, which handled the multitudes of daemons that were not attached to a primary character.

In the world of The Golden Compass, children’s daemons are shapeshifters, assuming the form of different animals according to mood or necessity. A decision was made early on not to try and use real animals on set – with the exception of a few dogs – but to create the daemons wholly in CG. The daemon of the main character, Lyra (Dakota Blue Richards), is Pantalaimon, often taking the shape of a cat. But Ray Chen, co-VFX supervisor at Rhythm & Hues, says the more difficult challenge was creating the golden monkey that accompanies Nicole Kidman’s devious Mrs. Coulter.

Chen worked on the lion Aslan for The Chronicles of Narnia: The Lion, the Witch and the Wardrobe, so he knows all about difficult animated characters. “In some ways, the golden monkey was a lot more difficult than Aslan because, although most daemons can speak, the golden monkey never has any dialogue,” Chen explained. “The performance needs to be relayed through facial expression, motion, and performance. And, as an animal, the golden monkey actually isn’t found in nature.”

Working with sculptures and concept paintings from the production, Rhythm & Hues built a sort of amalgam of golden tamarind, capuchin, and spider monkeys. The process was collaborative enough that the final monkey didn’t look a lot like the original concepts, but R&H spent enough time poring over the details of how real monkeys move their mouths and faces to create an elaborate rig that would make him as expressive as possible within the confines of a non-speaking character.

“It’s a small-sized animal, probably 12 to 18 inches high, and it has a large mane. The range of movement is a lot broader, more agile, than something like Aslan. The most Aslan would do is run. A lot of times he stood around and looked soulful. But with the golden monkey, there was a fair amount of interaction.” As an example, Chen cited an early scene in which the monkey attacks Pantalaimon in Mrs. Coulter’s apartment, which demanded strict attention being paid to fur dynamics. It all added up to a believably ferocious character that was tricky to pull off.

Another visual that required attention to detail was the quick transformations as the daemons change from one form into another. Based on early tests, it was decided that the change should happen quickly – over the course of a few frames – and would be done as a full transformation of the 3D shapes, rather than a 2D morph. “We have a number of shots with daemons transforming from short hair to longer hair, or from feathers into fur, so there was a fair amount of in-between stuff: feather-shrinking or fur-extending,” Chen explains. And then there were the daemon deaths – when a human dies, its daemon spontaneously dissolves in a puff of what’s known in Pullman’s universe as Dust. “We worked to create something that’s very fluid based,” Chen said. “We have an in-house fluid simulator called Ahab that we used to create gas and fluid simulations, and we used Houdini a lot for particle animations like snow or dust or footprints.”

Rhythm & Hues does some 3D modeling in Maya, but most animation, lighting and compositing take place in the studio’s long-lived proprietary software, including the animation and choreography package, Voodoo. For rendering, R&H uses its own Lighthouse lighting package with Wren, its renderer. The facility’s proprietary compositing package is called ICY.

Meanwhile, Cinesite was doing its own work in the fur-and-feather department, creating a small army of secondary daemons that would populate many of the film’s scenes showing humans gathered. “It was a challenge to integrate the daemons in the scenes,” says Matt Johnson, VFX supervisor at Cinesite. “They’re not just there as animals; they’re there to represent the actors’ personalities. So that was interesting from an animation point of view.”

Cinesite busied itself building photorealistic dogs, raccoons, even a praying mantis. A CG raven took full advantage of mission-specific refinements to the existing fur-and-feather pipeline. “Fur is challenging enough, but feathers are an order of magnitude beyond them,” said Thrain Shadbolt, Cinesite 3D supervisor.

“We have TDs now who could probably come up as fully qualified vets,” Johnson jokes. “They know an awful lot of things you wouldn’t really want to know about bird feathers.

Cinesite built its daemons using a combination of off-the-shelf software for animation and endering (Maya and RenderMan) and proprietary tools for fur and feather grooming and rendering. Cinesite composited in Shake, with some additional work done in Nuke, particularly using Nuke’s 3D camera. Particle effects were generated mostly in Houdini.

Neither human nor daemon are the story’s panserbjà¸rne – surly armored bears that figure prominently in the film’s second half. The task of creating those bears fell to Framestore CFC, where VFX supervisor Ben Morris was delighted to rise to the task.

“During the bidding process, I was completely shameless about the whole thing,” he said. “I read the book as I was doing Harry Potter 2, and I thought, ‘Jesus Christ – who’s going to get to do that bear? That would be fantastic!’ We laughed about it and carried on working. But when it came back as a potential project, there was such enthusiasm about it. We kept on telling them, ‘Give us the bears! Give us the bears!'”

The conception of Iorek Byrnison, the ragged bear who was cheated out of his rightful throne and lives in drunken ignominy behind a tavern, was critical not just because he becomes a major character, but also because he’s present in the film’s signature image, of young Lyra riding him at a fast gallop across the frozen tundra.

Framestore started out with concept maquettes of Iorek and his rival, Ragnar, and then Morris quickly set out to find more reference material. “There happened to be a trained polar bear at an exotic animal center just outside Oxford, so we went up there and spent a day shooting on two Panavision Genesis cameras,” he said. “And I happened to know there was a polar-bear skeleton in the Oxford University Museum of Natural History. I got the calipers out and took all the dimensional measurements.”

From those generic characteristics, Iorek’s character was developed. “Iorek went through a whole circle of evolution,” Morris recalls. “After the first one, everyone said, ‘No, he looks too scary. He’s big, he’s shaggy, he looks like a monster.’ So we slowly shortened the hair, reduced the clumping, and made him cuter and more princely. I have to admit, we were thinking, ‘Whoa, this isn’t the character we thought he was going to be.’ Luckily, it cycled back toward the dirty, longer-haired, down-and-out, grizzlier-looking bear.”

Because Iorek is a stoic, largely nondemonstrative character, he was well-suited to a fairly rigid bear face, Morris said. The villainous Ragnar, on the other hand, offered the opportunity for more anthropomorphization. “Rags is completely deluded,” Morris says. “He’s a schizo guy who’s desperate to be human. He’s got kind of an unhinged edge, so there was a little more scope for fun characterization.”

Beyond questions of character, there were the very practical issues of making the bears – full-fledged supporting characters in the story – believable on screen. “We knew these guys were going to come up and sort of rub their noses on the lens,” he says. “It meant the texture maps had to be of a certain resolution, the modeling detail had to be dense enough in the face to build the deformations and shapes that we needed. Everything would have to exist at that full-screen level – he’s one of those characters that didn’t really exist as an ‘easy shot,’ walking around in the background. Every time he turned around, he seemed to hog the screen.”

To reduce the rendering requirements whenever possible, Framestore was able to cut a few corners. Iorek’s physical geometry was consistent, but he had varying levels of rig complexity depending on the shot. In close-up, Iorek had up to seven million hairs. In a mid-shot, he had fewer than one million, thicker, hairs. And, of course, nothing that you can’t see from the camera’s viewpoint is actually rendered – Iorek is bald on his backside!

“Our fur system was actually rewritten for The Golden Compass,” says Morris. “We knew the complexity would require a more modular approach. It’s called FC Fur. It works on a stack of filters, so you have a filter that controls fur length, a filter that controls thickness, and a filter that controls scraggle. You have one that does curl and one that does clump. You stack them in different orders, and you can pipe the fur data in and out into cache files that allow you to save instances of static fur on disk. You can apply wind filters and collision-avoidance filters. You end up with a vast library of grooms for hero bears and guard bears. You have fight damage and scarring. They’re all different groom caches that somebody has to sit and make.”

The final trick to making convincing fur is paying attention to the details of lighting. On the shoot, Framestore captured HDRI images of each lighting setup; it also created synthetic HDRI images for each virtual environment. That data was then used in combination with traditional CG lighting techniques. “But that component of lighting is very important when you’re working with fur,” Morris explains, “especially with all the specularities and reflections you need in the fur.”

Framestore uses mainly Maya but also XSI for modeling (with the open-source Liquid plug-in for RIB generation) and RenderMan for rendering. Some of the artists work in Houdini and render using Mantra, which is a close match with RenderMan.

All the daemons and bears in the world are of no use unless you have someplace to put them, and the VFX teams on The Golden Compass spent lots of time working on virtual environments. In interviews, director Chris Weitz says he had hoped to shoot on real arctic locations but soon realized that idea was completely impractical. Green screens would have to do the trick.

Framestore had to create the environment for the film’s rousing polar-bear smackdown between Iorek and Ragnar, a complex task. “We were, in effect, making a photoreal CG movie for the duration of the fight,” Morris says. “We’ve got to kick up all of the snow in 3D around the feet, we’ve got to get snow onto the fur, we’ve got ice dust blowing across. It’s a very rich digital environment to be working in.

“But it peels down to a large sky dome, distant digital matte paintings, 2 ½-D projections in the mid-ground, and then full-on shaded 3D geometry and foreground FX elements.”

The shaders had to convey scatter and refraction inside the ice, and Fink had always envisioned the scene as a frozen ice lake with snow on top, meaning there had to be the suggestion of depth below the surface. “We render in RenderMan and generate a beauty that looks pretty good, but we can separate the shading characteristsics into arbitrary outs that we pass on to our Shake compositors,” Russell explains. “They can actually rebuild that beauty and change intensities, specularities, reflections and scatter coefficients. Because the renders are so intensive, we don’t want to be rendering them lots of times. We want flexibility on the back end.”

A custom snow-and-ice shader was used to procedurally generate distributions of snow and ice. “It would make sure snow settled on the top facings of the ice and rocks, and that you would reveal ice on the vertical edges. It also automatically generated smooth continuity between multiple geometries, so it would look like snow actually blended up to the ice blocks.”

Cinesite was largely responsible for the film’s climactic battle at Bolvangar, the forbidding place where kidnapped children are taken to be separated from their daemons. Matt Johnson describes the look of the prison-like facility as a Victorian, greenhouse-like combination of wrought iron and glass inspired in part by the architecture of London’s Kew Gardens and Crystal Palace. The scenes were shot at Elstree Studios with moving cameras on a green-screen stage covered with fake snow. The Cinesite team tracked the shots based on the camera data they could get from the live-action plates, and then created a full 360-degree world around them combining ground-plane CG snow elements and matte-painted mountain textures projected onto 2D geometry. The building itself was modeled in Maya and rendered in RenderMan using an array of custom shaders.

“Originally, the sequence was expected to be more of a matte-painted background, but as they cut developed, we quickly saw that we would need a lot more 3D,” Shadbolt says. “So we built, pretty much, a full 360-degree environment for camera placement. In the end, you had the ability to do some new camera moves and shots that were added very late in the production, and were entirely CG.”

Shadbolt describes a particularly complicated showcase shot that starts on Lyra, tilts up to a witch flying overhead, pans around to reveal the ground battle among hundreds of figures with witches flying through the air overhead, then closes in on the lead witch swooping close to the camera – actress Eva Green’s face was tracked onto the body of a stunt double.

“That was the advantage of the 3D environment,” says Johnson. “It literally spins the world around. That shot was a combination of live-action plates shot with a cablecam that were stabilized and rear-projected onto a new ground plane. We added hundreds of additional CG tartars and children running around in the background, as well as witches and arrows. We used some particle simulation around some of the daemons so they would kick up snow as they came to a halt.”

For Rhythm & Hues, the question of environmental work is a touch bittersweet – some of the facility’s most impressive work for the film was cut at the 11th hour, as the filmmakers rejiggered the book’s climax, essentially jettisoning the last few chapters’ worth of story. “It’s a shame, because a lot of the guys here worked long and hard on those scenes,” says Chen. “There were a number of shots of the North Pole collapsing, and huge ice pieces breaking off. We had a long R&D period to get a system up where we could actually have a landscape shatter and break into smaller pieces, as well as developing software for when the ice and snow crumbles, getting the pieces to interact and flow and have sort of eddies. None of the work is wasted, but it was a lot of build-up with no pay-off.”