Article Index

Narnia is a place where stars fall to earth to share their magic, mice are fencing champions and treacherous mists prey on the dreams of heroes. The Senate VFX, Cinesite and MPC describe their adventures through ‘The Chronicles of Narnia: The Voyage of the Dawn Treader’.


The Senate VFX completed 250 exciting shots for 'Dawn Treader', 215 of which appear in the finished film, covering full-scale CG environments to embellishments to bring some of the magic to Narnia.

The Senate VFX started creating previs for the opening shot in mid-2009 for VFX Supervisor Angus Bickerton to discuss with Director Michael Apted. The shot begins at a stained glass window high in a tower, pulling back and moving right down to street level – all created in 3D at The Senate VFX. As the camera pans the sky, a Spitfire flashes across the screen and they pull back across the road. The full shot was created in previs, for which they modelled the Cambridge King's College building and built a camera move. They took this move onset for the crew to try to reproduce using a Technocrane.

"The crew wasn't able to achieve the first part of the shot, but we did manage to create moves that simulated the way a crane moves in terms of tilt, pitch and height, and stitched the live action camera onto their 3D camera," explained VFX Supervisor Richard Higham. Consequently, the first part of the shot is the invented camera, and the second part is live action, still displaying our 3D buildings in the background. The only real objects in the scene are the people crossing the road, the road itself and the railings. The buildings are all CG – the gatehouse and walls, King's College chapel and other college buildings – requiring extensive detail and time."

Turning Blue
Ideally, Richard always prefers to work with HDR data for lighting but in this situation, the production used about 140ft of blue screen, which effectively turned all on-set lighting blue. So they had to apply some creative lighting to produce the correct time of day.
Richard said, "Cambridge, of course, was extensively photographed at high resolution to use for modelling, colour and to extract textures. But the college is a very large building and they had no access to the upper portions, requiring long lens zooms and manually calculating the correct dimensions, as well as adding in realistic detail, breaks, chips and weathering. We were also building our work to come very close to camera, so all CG had to be high-res from start to finish. A quick projection wouldn't work."

They were able to follow the previs'd camera move right down to where the real camera began, shooting people on the street. The idea was to let the audience wonder for a few moments whether they were already in Narnia, revealing only the colourful stained glass window, stone structures, spires and ramparts. But the Spitfire was their first clue of England, also created by The Senate VFX team.

Unfortunately, no architectural layouts or drawings for Cambridge University were available, which is why they needed so much photography. Michael Apted had spent some years in Cambridge and wanted a very accurate result. They had to calculate positions from multiple photos and correct distortion in others shot with different lenses, using tools in Photoshop and their compositing software.

Bell Tower
Early in the story at Narrowhaven, the main city on the Lone Islands, the heroes enter a dark, cavernous bell tower. From high above in the ceiling, a crew of dangerous characters suddenly descends on ropes and attacks them. "Nearly the entire environment was built in CG, apart from the floor and several statues seen in the background," said Richard. "The upper reaches of the tower and bells concealing the attackers had to be created. The production wanted to see basilica style architecture over and around them, but the plate only provided a black space, showing no scope. Texturing, detail and appropriately styled arches and pillars behind the statues were required, combined with a great sense of height."

The artists based their model on another they had built for 'The Da Vinci Code' for the Council of Nicaea shot. The team handed over the model to the Art Department to experiment with and modify, and when it came back, they touched it up and textured it, and added shaders.
"Then we could start on the lighting. Again, we started with HDRI but had to creatively interpret lighting for the newly built portions. Angus got involved and wanted more light through the background. Atmospheric beams of light need volumetric lighting that, as far as possible, has to match light sources from the live action. Set lighting creates its own volumetric atmosphere. The team tries to blend any new lights but every scene needs to be handled shot by shot, based on the director's request and artistic decision.

"In this environment, I think the decision to use a full 3D environment was smarter than adding 2D elements because we could give it more scope and a sense of depth, including light. Of course, without blue screen, everyone had to be individually rotoscoped out throughout the sequence and the new background placed behind them."

Interior Weather
When Lucy Pevensie enters the enchanted house of the magician Coriakin, she finds herself in his library where she reads from a book of magical spells standing on a lectern. As she blows the dust off of it, a jumble of letters on the cover suddenly animate and rearrange themselves into the book's title, 'The Book of Incantations'. As she opens the book and reads one of the spells, a snow flake drifts down onto the page – added in 3D CG.

From there, the camera shifts into a reverse angle looking up at her, and when it changes back again and she looks up, a completely 3D CG environment of the library itself is revealed, this time with falling snow and clouds, resulting from the spell she read out that has brought the weather system inside. The entire library was built and modeled in Maya, based on the production's working 3D model.

This model was received as a Maya .obj file which the team detailed and made more accurate with texturing and tonality to match the surrounding shots. As soon as the artists start increasing complexit and adding photorealism to a working model, the mesh usually needs to be reworked. HDRI lighting from the set is also applied, and in this sequence it was especially beautiful.

Snow Storm
"The main task here was adding the snowfall," said Richard. "Purely due to the angle of the shot – looking up toward the ceiling - a single particle system wouldn't suffice. We had about seven different particle system layers, using only one 2D foreground element to place in front of Lucy plus an area excluding the background snow for her body, to run the particle simulation.

"We created a range of snowflakes, fairly simple in design, that wouldn't come too close to camera, and also needed a selection to match the practical flakes from the snow blowers on set in the surrounding shots. These were used in various simulations – one placed above, close to camera falling directly toward viewers, then a mid-ground simulation falling behind her, a background set, others placed higher to fill other areas. They were all created from those 20 or so snowflake variants, but because of their random movement, they seem more varied."

The lighting, based on the HDR, would be adjusted according to the shots on either side, tweaking the colour and direction. However, this lighting rig then had to be adjusted creatively to work with the snow fall, adding light to different areas or some strafing beams for the flakes to pass through.

Clever Pages
In the case of the book, they had the physical prop to work with, to photograph and model the main parts. They aimed to recreate as little of it as possible. When Lucy opens the book, they wouldn't need to recreate the leather binding which is hidden, for example, but only 80 to 100 of the pages - those that begin magically flipping over by themselves.

"Each page is basically a plane requiring texture. The design was derived from the original artwork used to create the physical book, applied directly to their geometry, to which we added a paper texture plus staining to age the page, and finally lit to match the specific lighting from an overhead source. The pages were then rigged for simulation. Although this would need a lot more processing and less individual control, a more realistic movement could be achieved than attempting to animate the geometry. During the simulation, it had to flip and cast shadows on itself, so each page had to be properly lit and shaded for a more photorealistic look."

At one point, as the pages begin to turn, Lucy panics and slams her hand down to stop them. The Senate team recreated that moment of interaction as the page folds under her arm, and rotoscoped her arm back over the top of the CG. It required them to figure just how much CG flipping was needed between cutting in and out of the live action hand movement.

For the title's letters, they needed two front cover designs from the Art Department. The first showed the letters scattered in random order, the second with the letters in their final position showing the title. They took the first instance and worked out a 2D previs animating these letters, aiming to end up with the second design. They experimented until they were satisfied and showed this animation to the actress on set to explain what would happen when she blew on the book, and help her perform accordingly. They also made sure the lighting and textures of the moving letters matched the real letters, which had glints and indentations in them.

Octous' Library
Shortly after this, Coriakin the magician himself literally emerges from the background of his library. Angus admired the way an octopus can appear to 'peel' itself out of a background as it discards its camouflaged looks, and wanted Coriakin to suddenly emerge from this sequence's environment in the same way.

Richard and the Senate team devised some 3D techniques for doing this, simple in theory but complex in execution. Because it had to be planned carefully, the Senate team opted for a motion control shot using a dolly tracked toward a table at the back of the library, holding glass jars, books and plants in front of a large bay filled with more books. Tracking toward that arrangement, they captured two takes. The first included the actor, who turns and moves toward camera. The second take was captured with an identical move but without the actor.

Starting with a clean frame of the background, Coriakin was re-modelled to match his contours and the clean frame was projected onto a 3D version of him. Richard explained, "If the camera had been standing still, you would have seen nothing, but as it travels forward, the model no longer perspective-shifts in the way that your eyes expect. You notice an odd change as the background peels forward, which the artists enhanced by modelling refractive glass and other textures into the Coriakin model. As you get closer still to him, those details become more obvious, changing into a defined shape."

Finally, they used a matte created with a 2D file sequence projected onto the actual texture, producing an ink-bleed effect that reveals the live-action through this matte, giving an illusion of something moving toward you, bleeding into the real Coriakin. "You could say we pulled magic from nature," said Richard.

Animated Geography
By the time the magician Coriakin appears in his library, all the heroes have gathered with Lucy and he directs them into the next stage of the adventure. He pulls out a scroll and throws it out toward camera, when it unrolls onto the floor and a number of islands and clouds appear as on a living map, all of which had to be created in 3D CG and composited over an interesting arrangement on set. A dark blue screen, about 18ft x 9ft, was unrolled and flat, bright white lights were placed onto it to create an uplighting effect onto the actors. These had to be painted out.

As Coriakin talks to them about the map and he walks over the water and among the clouds, his feet interact slightly with the surface, which is slightly resilient like as LCD screen, and the clouds swirl and wisp a little around his legs. The only way to introduce their work for this map was to rotoscope him out of the footage.

The map consisted of the ocean, the islands at different scales, and fluid simulations to create the clouds. Fortunately they didn't need to create water interactions, only the LCD effect. The islands are animated throughout the sequence – the map moves along for the viewer rather the way a microfilm search machine moves to and enlarges an area of interest. As Coriakin described a location, such as Ramandu's Island, the map would zero in on it as it came to life. This was achieved by flying the camera low over the ocean toward the island as if the viewer was approaching it, but as the camera came up over the top, he realizes that it is still just a map in a room. Dropping down through the clouds added a similar stylistic touch, at which point they handed the shot over to another facility to finish the progression.

Camera Matchmove
Camera work required 3D tracking the live action, starting with a scene and the real camera. Within that, they had to create the actual movement of the islands on the map as well, requiring their own coordinate system which moves within the map frame. "This is really just tracking the live action. On set, the crew planned to keep their camera low for our team, knowing that the 3D would eventually cover up what was actually there, until the camera comes up over the top of the island," Richard said.

"We were using the live action camera, but on all 3D shots we created our own cameras for the projected environment, matchmoving back to the real camera. In fact, every time we put 3D into a shot, unless that camera was fully locked off, they needed to matchmove the cameras, ensuring the parallax, perspective shift and other factors are identical. The 3D camera has to be a matchmove camera solve from the live action."

Falling Star
After previs in July, Richard was on set in September 2009 and scans started to come through in November and December, but they could start some things in advance, such as the huge environment background for Ramandu's Island. They built trees and ruins for the island that they knew would be required and could adapt then the plate arrived.

"This island contained several important sequences for us. When the character Liliandil transforms from a star, dropping down to the island, the effect results from particle simulations we created to resemble images available to everyone from NASA. Because now people know what a star looks like, we wanted that kind of realism, not a glowing child-like version of a star with solar flares and swirling plasma making her an organic real star. A couple of months of R&D on the particle and fluid simulations were required, using Maya particles and fluids, experimenting with textures, gravity, movement and flow limes.

"We ended with five to ten different simulations emitting from a central sphere which we applied to a model derived from a scan of Liliandil that we made our basic mesh from. We placed this where she would be standing to match the contours, and used this same simulation – but emitted from the new Liliandil model without rendering the actress herself, just the emission."

The viewer understands that that someone will appear in that shape, before the live action is blended into the emission, when she appears to be emitting the particles herself. This kind of transformation repeats as she reforms into the sphere, throwing off particles and gases similar to a super nova to shoot back up into the sky.

Perfect Environment
The camera then jumps back into a reverse angle revealing a full CG environment, because there was no photography of the island environment at that angle. Similarly, when the ship arrives at the island, the viewer sees a beautiful sunset with unusual island formations, created with matte paintings based on library photos from the English Lake District worked onto an ocean plate from the production.

This could be accomplished in 2D but in other views, such as in among the trees, waterfalls and aqueducts, had to be created and embellished in 3D. Extensive background vegetation with strafing volumetric beams of light was also built and could be used through the entire sequence, but almost every shot needed bespoke modelling of its own. The characters return to the island later as well, which made perfecting the environment important.

Tricky Mists
An evil green mist, surrounding the Dawn Treader as she enters the treacherous waters around the Dark Island, takes the form of tendrils that play tricks on the character's minds and take the shape of their greatest fears and desires. This mist was one of Cinesite's principle tasks on 'Narnia', required to suggest evil and move in a distinctive style. The team decided that the best way to meet these demands and match the description was to first create actual objects that they could animate and assign looks to. Thus, the team's main challenge was giving character qualities to these mist tendrils, which are essentially fluids.

They built and animated what they called vicious-looking 'eels' to crawl through the scenes. The task then became translating the animation to the fluid simulations in a way that retained that same creepy, snake-like look so that viewers would get a sense of danger and evil, expressed through their movement, pauses and turns. The animations were cached and the fluid simulation was derived from this data.

"The simulation itself was created with Cinesite's proprietary fluid software developed in Houdini, called csSmoke. Created for previous films, this was an opportunity to push it further and use it to add character qualities to smoke objects without eyes, voices or faces. Each had three fluid simulation layers and 15 render passes for the lighting including internal glows and three different shadow passes," said 3D Supervisor Stephane Paris

Interactive Smoke
On top of these looks, the fluid renders had to react to their environments, demanding considerable matchmoving. To set up these interactions, animated digi-doubles were used for several of the characters, and some of the lighting elements, such as torches on board the boat, had to be animated to work properly with the smoke. Because it was semi-transparent, some light would pass through, and some needed to react inside of it.

One of VFX Supervisor Matt Johnson's favourite effects was when the smoke was to pass through a vent or even a solid bulkhead on deck. "As it is hits the surface, a slight displacement of the smoke occurs, as well as a portion that passes through. Creating effects in this manner helps ground them and embed them into the scene better than just having a hard rotoscoped edge," he said.

The speed of the animations affected the look of the simulation as well. Accelerating too quickly, for example, could 'kill' some aspect of the smoke's movement, so making the smoke look consistent from shot to shot was not straightforward. For animation reference, realistic smoke action didn't work as well as more mysterious, spooky dancing motion, similar to the looks that effects teams formerly achieved with cloud tanks. During look development, the production wasn't completely clear about the looks they wanted, or even the colour, which meant considerable testing and looking at references from previous films as far back as 'Raiders of the Lost Ark'.

Chunky Renders
The need for rotoscoping arose because as the camera moved through the set where the mist was to be created, it would often capture perhaps 50 actors on deck standing among large props, some of whom had to appear in front of various smoke layers and some that needed to remain behind them. Whichever figures the smoke had to pass around or in front of, the artists had to rotoscope out and use to create a matte to place the original photography back over.
The render mainly became complicated when composing the shots together. As the camera moved down the length of the boat, for example, they had to include the blue screen live action plate, all the associated roto layers, the digitally created background, CG water, the tendrils of CG mist and the 15 light and shadow passes they required, glows on the flambeaus and torches, plus any atmospheric mist and fog within the frames, all in separate layers requiring their own roto for objects at different depths. "So these were quite chunky renders – but exciting for team at the same time. Everyone knew already that it looked cool," said Matt.

Murky Environment
All on-board sequences were shot with the actors on a 360° blue-screen, without water or even a complete set for the boat. Many such elements had to be created digitally. Imagery and data for different backgrounds used in the movie was supplied from different sources but, although a practical boat set was shot in Queensland, no actual background images were available to be scanned for their 'mist' sequences around the Dark Island.

Therefore, they created their own dark, murky environment based on descriptions from the production, giving it colour, creating geometry to define the topology and adding rocks. They also worked with the camera positions to establish continuity from shot to shot. Once they had completed this environment, they shared it with other vendor's teams working in the same part of the story so that all shots would fit together.

A green pulsing energy, which they called 'synaptic pulses', was required in the smoke and animated as glows in the background. A further interesting aspect of the backgrounds, essentially 3D environments established in CG, was that the team could handle the matte paintings, 3D geometry and various elements within the 2D compositing stage. Then, using the 3D capabilities of Nuke, they could blur the boundaries between traditional 2D and 3D effects, which they found gave them a greater range of options for handling the shots.

Misty Witch
The White Witch character, another task assigned to the team, was in some ways quite a different kind of effects challenge, but in the story she results from the mist's ability to embody people's fears and confuse desire with reality. Stephane said, "Consequently the team had to establish a connection between the two entities – the witch and the mist. She would also require fluid and dynamic simulations but fortunately, we also had blue screen footage of actress Tilda Swinton's performance captured with her wearing tracking markers, and a 3D scan of her body. We could use these as the basis for the CG Witch, unlike the mist eels."

They were able to 'roto-animate' or match move this 3D digi-double to emulate the Witch's performance on set, frame by frame. That, in turn allowed them to digitally create and then animate her hair to give a flowing, misty look – all derived from the 3D maquette, which Stephane's team rendered with the required passes. They could use these hair and body animations to produce their own distinct kinds of smoke.

In the compositing stage, they worked on the look of her face, softening and smoothing it. A 'pro-mist' version of the face was created using a camera filter to take away lines and edges, making her both more ghostly and ethereal, and her performance more tempting and seductive. Some parts of the face faded out and back in, like the smoke, but they always focussed on the eyes, adding a green tint to them. Her internal illumination was rendered in the same way as the mist. In the end the viewer is not certain if she was a real ghost, come to torment Edmund, or purely a misty figment of delusion.

Glittering Bling
Footage for Goldwater Island, where the characters discover a jewelled valley, was shot in Queensland but only the set immediately around the characters was real. Cinesite created digital matte paintings for the valley, projected onto Lidar scanned geometry, and also created real CG rock geometry to extend the view with unusual, fantasy formations. "W also made the gold captured on set gleam and sparkle in the sunlight with our own digital 'bling',"said Matt, "combining grading techniques for brightness with many small, individual anamorphic lens flares composited and tracked in as little glows and glints. Then we added caustic render passes creating rippling light along the rocks and walls of the jewelled valley, subtly revealing how such a place would really look if lined with glittering jewels at the bottom."

At Narrowhaven, all buildings extending up a hill into the background are part of a CG environment the Cinesite team created with Moroccan-style architecture. Layout and placement of buildings of varying designs and looks were accomplished with Cinesite's Citybuilder software, used extensively in 'Prince of Persia'.

However, the artists had to ensure that the houses had an organic, naturalistic appearance, not precisely geometric, as if built of worn mud brick with imperfections. They made a 3D layout for the city over the island and for wide shots could project digital matte paintings over geometry, but they built more detailed CG buildings for closer shots in among the street, modelled with Maya and Mudbox, textured with Photoshop and Mudbox, all composited in Nuke.

Of Mice and Dragons
MPC completed over 700 shots in 'Dawn Treader' many of which involved two animated CG characters viewers know from the previous films, Reepicheep the dashing mouse, and Lucy and Edmund's cousin Eustace, transformed into a glittering dragon.

"We had designed and did the concept development for Reepicheep at MPC, beginning on 'Prince Caspian' and going through a series of ideas during pre-production on his look, colouring, costume and the props he uses," said Greg Fisher, Head of Animation at MPC. "For 'Dawn Treader' we took this original design and tweaked him a little further. Having created his asset from the ground up helped us understand how to get more from him this time.

"We really wanted to make sure that his moves and swashbuckling adventures with his sword had a basis in real swordplay. So it was critical that he hold himself and move correctly. In early pre-production, the whole animation team attended fencing classes to understand how to carry a sword and use it properly. While It is something that you can be told or watch how to do, there is no substitute for actually doing it. The animators could feel the weight of the sword and understand the body posture and weight shifts needed to create a move effectively. It turned out to be invaluable experience and we had fun doing it, too."

Mouse Maintenance
Nevertheless, they also used a lot of mouse and rodent reference to maintain a mouse character and the key aspects of being a mouse - with a little creative license, of course. For his human side, they used video reference of the voice actor Simon Pegg, as well as the action and characteristics of the classic movie swashbucklers like Errol Flynn. The skill, attention to detail and actual amount of anthropomorphism to add was the challenge.

Greg said, "How much of a mouse against how human he should feel was quantified shot by shot depending on what he was doing. Pure acting shots were more on the human side, those with locomotion tended more towards the mouse. The trickier shots had a bit of both, where we aimed to giving him the twitchy, nervous, high-frequency motion attributed to a mouse, but keeping him strong and bold in keeping with Reepicheep's character.

Second Chance
The team found one of the best things about working on him again was the chance to not only tweak his model and rig to age him but also to take what they had learned the first time and make him better to animate with, resulting in better final shots. The main feature they changed was his facial design. "You'll notice if you compare him between the two movies that his eyes have been brought further around to the front of his face, making him more human or predator-like, compared to more of an object of prey with his eyes on the side of his head, not to make him into a predator but to help with eyelines and acting shots.

"It had proved difficult on 'Prince Caspian' to reach a good, human type of expression when he always had to be slightly on profile while speaking to someone in directly in front of him. He could become more expressive and emotive, as this movie demanded of him. To age him we 'tubbied' him up very slightly around the waist region to look just a little portly, and a touch of grey was added on the whiskers and in the groom of his fur." MPC's work on Reepicheep has been nominated in the 2011 VES Awards for Outstanding Animated Character in a Live Action Feature.

Back and Forth
Tackiling Eustace the Dragon, Animation Lead Benoit Dubuc explained that by definition, a dragon rig is fairly complex, which means that when an animator approached a shot, he would try to plan it out as thoroughly as possible, mostly in terms of which 'space' to use for the various controls for his head, wings, tail and so on, and whether he would have to switch them around during a shot. "Preparation is an animator's best friend," he said.

"That being said, what we wanted to achieve with the Dragon rig was something that would allow the animators to focus solely on the performance and not have to constantly fight against an overly complex rig to get their vision on screen. The only way to achieve this is with a healthy back and forth relationship between the rigging and animation departments."

When the riggers handed over the initial rig to the animators in pre-production, they tested it out as thoroughly as possible. Then the rigger would address the notes that came from the tests and they would repeat the process a few times. One of the trickiest parts to get right was the neck, both to rig and to animate.

Exploring the Rig
"The first Dragon sequence my team worked on was the one on the beach, which was fairly challenging and a good opportunity to fully explore the rig," Benoit said. "The challenge came from the dragon sitting nearly front-on to camera, upright on his hind legs, and having to animate the wings interacting with the ground. The camera angle frequently foreshortened his neck, and his sitting position highlighted further aspects of the rig that needed ironing out, while the necessary wing-ground interactions required us to take advantage of the space-switching options on the rig.
"It was a difficult but good sequence to start with because after a bit more back and forth, we ended with a very good, efficient rig. And luckily for us, the rigging department was very accommodating to the animators' needs throughout the show, sometimes even to shot-specific requests."

The animation was not divided up into character specific teams. Instead, they had three teams, each assigned specific sequences. Within the sequence, they split the shots into beats that were assigned to the animators. However, as the show went on, certain animators developed a genuine familiarity with the Dragon character and were often assigned his more complex shots, which made it much easier to keep his character consistent.

Character Arc
Because the design for the Dragon's face was inspired by the actor who plays Eustace, Will Poulter, the facial rig was based on a human facial muscle system. In terms of animation, they drew inspiration from Will Poulter's performance. "His character arc occurs while he's in Dragon form, which was a great challenge!" said Benoit. "At the start of the film, he's an obnoxious, annoying character and by the end, he's been humbled and is much calmer and wiser. Both ranges of his performance were very useful to us as we had to transition, over the course of the film, from one performance to the other.

"At the start of each sequence, the entire team would get briefed on the intention of the sequence, the character arc that we were hitting throughout the sequence and the character's inner thoughts, especially important with a mute character. Then, on a shot specific level, the animators would often film themselves acting out the performance and use that as reference for their characters. This is the best way to add subtlety to your work."

Words: Adriene Hurst
Images: Courtesy of Twentieth Century Fox

Featured in Digital Media World. Subscribe to the print edition of the magazine and receive the full story with all the images delivered to you. Only $79 per year. 
PDF version only $29 per year