They can be found in every society, con men who take advantage of others while capitalizing on their victims’ weaknesses and fears. In real life, their deceitful deeds are no laughing matter. But in the world of make-believe, their actions and consequences can be hilarious, and the characters themselves endearing. Such is the case with Gwizdo, a medieval swindler who stars in Dragon Hunters, a European CG feature film.
Dragon Hunters follows the adventures of Gwizdo and his partner, Lian-Chu, both hunters for hire in a futuristic/medieval world of floating land masses terrorized by various monsters, also known as dragons. In this timeless universe, dragon hunting is not a fairy tale, but a real job involving contracts, money, and problems with clients who don’t want to pay. Knights work for free but are not always available when needed, thus providing opportunity for commercial hunters.
Small in size, cowardly, and less-than-honorable, Gwizdo is the brains of the operation, negotiating contracts with the helpless villagers, while the hulking but kindhearted warrior Lian-Chu provides the brawn (it is his task to slay the beasts). However, that seldom happens. Both pretend to be brave and collect advance payments for dragon kills that never occur. While opposites in stature and personality, the two have been longtime friends, having been raised together in an orphanage. Often, they are accompanied by their pet dragon, Hector, which is more like a pet dog than the large, menacing dragons they hunt. Then, one day, they meet little Zoe, a believer in fairy tales and the niece of a wealthy nobleman frightened by the return of the fiercest dragon of them all, the World Eater. Soon, the fly-by-night dragon hunters find themselves on an exciting adventure: their first actual hunt.
The underlying story is universal in nature, a description that also holds true for the multinational group of nearly 450 people behind this film. Mac Guff Ligne in France, Trixter Film in Germany, and LuxAnimation in Luxemburg were responsible for the movie’s creation. French graphic novelist Arthur Qwak developed the original story and served as co-director alongside Guillaume Ivernel, with the French company Futurikon producing. The movie found eager audiences throughout Europe but is still waiting for possible North American distribution.
Nevertheless, some folks in the US may recognize the story line and the characters. That’s because the property began in 2004 as a television series on Cartoon Network. “From the beginning, Dragon Hunters was conceived as a concept that could be a series, a video game, a feature film, and so forth,” says Caroline Blin, marketing director at Futurikon. To that end, Futurikon recently completed a second season (52 half-hour episodes) of the series and rolled out a video game for the Nintendo DS to coincide with the film debut.
Even though the concept was imagined as both a feature and a TV series from the start, one of the biggest hurdles was getting the producers to pony up the necessary funds for a 3D film infused with a tremendous amount of effects. “We were like Gwizdo and Lian-Chu: two lonely rogues who needed credibility!” jokes Qwak.
The two con men-turned heroes face their biggest fear, literally, in the fi lm as they chase down the fierce World Eater, whose immense size (450 feet tall) challenged the hunters as well as the artists.
Film Production
Blin is quick to point out that the movie—a prequel to the series, a “how it all began”—is more than a mere extension of the TV shows in story, tone, scope, and character development. The biggest difference, though, is the animation itself: The series has a traditional 2D cel look, while the film is 3D CGI. Therefore, no assets from the series could be reused for the film. Yet, the 3D style is atypical of the CG films produced in the US, with their rounded edges, glossy imagery, and tighter frames. VFX producer Jean-Jacques Benhamou from Mac Guff describes the fairy-tale style of Dragon Hunters as “a Miyazaki mood mixed with German romantics influences.”
In contrast, the worlds of Dragon Hunters contain realistic surfaces (grass, water, stones, fire) but with unique physics in which everything is moving all the time. “The worlds had to feel real in order for the adventure to be credible,” says Qwak. “So we approached it similar to a live-action film. That meant realistic rendering, matte paintings, and a tremendous amount of work on compositing. The characters, while cartoon-shaped, have realistic textures.”
According to executive producer Michael Coldewey from Trixter, the shape of the characters and the entire production design is one of immense contrasts. “If we have to show something big, it is big in this movie. If there is a happy moment, we show little, fluffy, white bunnies flying through the sky,” he adds. Coldewey also believes the film gives audiences time to enjoy the artwork: “We have some fast-edited sequences and action, but we also provide emotional pauses to relax and enjoy the art.”
Most unusual are the film’s environments, which are in constant motion and float in midair, as do elements and debris, created with a particle-based system and turbulence field.
Character Building
In all, the film took more than two and a half years to complete—one year of development (storyboards, designs, references, and a 2D animatic) and another year and a half of CGI production. The majority of the film’s CG work was done by Mac Guff, which created the characters, rigs, textures, and some animation. The facility also did all the background renderings and composites of the final images. Trixter, meanwhile, was responsible for the character animation and the musical score. LuxAnimation modeled the props, sets, and backgrounds.
In addition to a sometimes language barrier among the three studios, the facilities had to ensure that their production pipelines were connected and compatible for asset sharing. Also, the studios had to work as a single production unit.
Using sketches, paintings, and clay models as references, the crew at Mac Guff used Autodesk’s Maya, along with a good amount of in-house software and special scripts, to bring the Dragon Hunters characters to 3D life. Character layouts soon followed to validate the 2D-to-3D transfer in terms of the look, proportions, and so forth.
One of the biggest challenges, says modeling and lighting supervisor Nicolas Brach, was creating some very stylized characters that would be integrated in very realistic environments—a look they dubbed “illustrated realism.” This required the modelers to focus on adding fine details to the models while retaining their “uncluttered style.” Often, this meant adding more modeling detail to the character’s props and suits rather than to the cartoonish heads and faces. For example, Knight Lens Flair and his horse have very realistic, detailed armor on more stylized and cartoonish body structures and proportions.
While all the characters had their unique qualities, the most difficult to model, says Brach, was Hector the blue dog-dragon—not because of his shape, but his topology. “From the beginning, we knew we had to consider all the motion possibilities he should have,” he says. “We had to be able to squash and stretch all parts of his body, including his eyes and teeth, without crashing the skinning and blendshape process.” Another was the World Eater because of its size (450 feet tall); also, it was to appear in both wide shots and close-ups.
For the hair and fur—from Zoe’s braids and Lian-Chu’s ponytail, to the fur trim on the clothes—the artists at Mac Guff used the studio’s proprietary tool called Symbor. With the tool, they generated guide splines on each vertex of the fur, then “brushed” them into the modeling layout with special tools, akin to using combs. Afterward, they created a large number of hairs to fill the surface using those guides. For every character, the team had at least three hair systems for different densities and thicknesses, and then chose the most appropriate system based on the location of the character on the screen and its proximity to the camera. “We didn’t want to unnecessarily overload the memory,” Brach adds.
The hair and fur—as well as the clothes—moved with the help of dynamics, as well as with Syflex’s cloth-simulation tool and the plug-in from Mac Guff.
For the most part, the group hand-painted the textures in Adobe’s Photoshop and used Pixologic’s ZBrush as well to create displacement maps and deep UVs to deal with the film’s complex texture projections. To ensure that the stylized characters fit well aesthetically into the backdrops, the artists used realistic textures, acquiring basic materials from photographs and then mixing the textures in Photoshop. They used this process to create the clothes, leather goods, metal objects, skin, and other materials.
With Symbor, the artists then built complex shaders driven by mathematical parameters such as facing ratio and distance to the camera. They used these Blinn shaders for basic surfaces such as leather and metals, some velvet shaders for clothes and horse hair, and subsurface scattering for the skin.
“We had a lot of subtle shading details on things like Zoe’s clothes or her Uncle Arnold’s clothes: Both wear garments embroidered with gold and have different velvet shaders to give them a sense of fineness and richness,” Brach says. Nevertheless, the artists used what Brach describes as a “fat S” shader for the hairs instead of a traditional hair shader, for a soft look achieved with a short render time.
Lighting and rendering was done with Mac Guff’s Symbor as well, combined with MGLR, the studio’s rendering engine. The light rig was based on ambient occlusion, which allowed the group to control the mood of the sequence; sometimes image-based lighting was used, along with some raytraced light sources.
Mac Guff’s R&D department created a multilayer system based on various outputs that enabled the group to generate basic layers (occlusion pass, diffuse pass, normal camera pass, and so forth), as well as the possible mask and shader output needed by the compositing department—all in a single process. “We even rendered a layer for each light source, with all the possible information (diffuse, spec, shadow) so we could refine the lighting color in real time within 2D,” Brach says.
Animated Characters
After the modelers were finished, the files were handed off to the rigging department. Each biped character rig—set up in Maya and managed with InK, the studio’s in-house tool—contained approximately 75 joints, with 100 or so controllers. “Because this is cartoonish animation, everything had to be squashable, stretchable, bendable,” says Mathieu Trintzius, character rigger, “but I wouldn’t say that the animation falls into the cartoonish style; it is more realistic animation, with the exception of Hector.”
The characters, including the hulking Lian-Chu and little Zoe, are clearly stylized, yet to fit them into the realistic environments, artists added fine details to the characters’ clothing and props.
Even some of the dragons were rigged with the base system, including the World Eater, although the group devised special rigs for creatures that had unique designs. This also held true for the props.
All the characters were then hand-animated in Maya, reflecting a style that animation supervisor Kyle Balda describes as “somewhere between full and a hybrid ‘limited’ animation in 3D”—hardly surprising given Qwak’s traditional animation background. “There is a lot of sappy pose-to-pose animation, but for the most part, the style of the animation is driven more from the personality of the characters,” explains Balda. “For example, Lian-Chu hardly moves except for when he is in an action sequence, so his acting is subtle compared to Gwizdo, who moves a lot as he talks, but more or less freezes up when it’s time for action.”
According to Laurent de la Chapelle, animation director, the team was tasked with creating a full feature’s worth of animation in about nine months, and with 26 animators divided between the Mac Guff and Trixter facilities—about half the time and resources as a Hollywood animated feature.
The groups commonly employed video reference, which enabled the directors to communicate the acting/action intentions for a sequence that was being animated long distance by Trixter in Munich. “This was simply a way to communicate the character’s motivation for a sequence, as opposed to using a detailed guideline of the exact poses we wanted,” notes Balda.
The facial setups were based on muscular movements in the face, rather than the creation of shapes that reference particular phonemes or emotional expressions. In this way, the animators were not limited to the same “F-shape” or the same surprised eyebrow pose. “It was up to them to sculpt the poses themselves based on a combination of the anatomy of the character and their own graphic sensibilities,” Balda says.
Each character model contained several facial controllers on the eyes and mouth, giving the animators a good deal of freedom in creating facial expressions. This system negated any need to use facial poses. No motion capture was used, though some characters were influenced by real-life actors—Gwizdo (Joe Pesci) and Zoe (Kate Hudson from Almost Famous, and Miranda Cosgrove from School of Rock).
Another challenge was that, due to the production schedule, the teams began animating before the final dialog was recorded, limiting them to only animating portions of sequences that were not dialog-driven for the first few months. “Fortunately, there were plenty of action sequences in Dragon Hunters, but not all the animators are necessarily differentiated toward action-based animation; some are natural actors who may struggle a bit with the concrete and technical nature of physicality,” explains Balda.
In addition, casting was still going on in the early months of the film, but that limitation proved beneficial when the crew moved to the acting sequences: “By that time, the animators had become familiar with every joint and anatomical movement of the characters,” Balda points out.
Lip synch proved especially challenging, as the dialog was created in English to attract a widespread international audience, while the animators were predominantly French. To make this process go more smoothly, Balda accompanied the directors to the voice-recording sessions in Los Angeles, to offer a native perspective. “For an animation supervisor, this is an unusual practice, but now I am convinced that it should be part of the process,” Balda says. “Within eight days we recorded all the character voices, focusing on one character at a time, from start to finish, from the perspective of that specific character. Later, this proved invaluable when I was working one-on-one with the animators to help them understand where their character was in his or her respective arc.”
Balda also gave the animators a video reference of him lip-synching the final recorded dialog, so the French animators had a reference of the mouth poses that were not obvious to a non-English speaker—this was particularly helpful for Gwizdo and Zoe, who talk very fast and use a lot of slang.
Environmental Issues
One unique aspect of the film is the floating islands. “We had to take into account the fact that all the animations were made before the modeling of the high-definition sets even began, and that the animators worked on low-definition scenes made solely for them,” explains Franck Clarenc, modeling supervisor working on the film’s sets and environments. “As a consequence, all the assets in our pipeline were created according to that low-definition layout, which was divided into arbitrary parts that were much too heavy when turned into high definition.”
Artists surrounded each set with a sphere mapped with a high-resolution matte-painted sky. Most of the reflections are actual blurred reflections; sometimes they are environmental maps.
In the case of the floating village, the artists created it as a single scene, which, according to Clarenc, turned into a huge liability when they reached the critical polygon mass that made working in real time within Maya “a distant memory.” So, the group cut the huge data into smaller bits, and with help from the development team, automated the gathering of those sub-scenes into a single, massive one, thereby enabling the artists to work on the various individual parts in a more convenient way.
“Basically, we visualized in our asset manager the sub-scenes and the global scene in the form of a nodal tree, where we could connect, disconnect, and replace at will any part we wanted,” explains Clarenc. “This was a major achievement in flexibility, work division, and efficiency for us.”
Throughout the film, there is a lot of debris flying around, some of which are just rocks, but others are part of destroyed buildings. To achieve the “floating” effect, the group devised a particle-based system to generate a turbulence field; that, along with a direction-based rotation script, moved the “flying” rocks, vegetation, and more.
There are also a number of buildings in various states of disarray during different “chapters” of the story. Because of the tight production schedule, the group had to find a way to optimize the creation of those heavy models; in the end, they designed most of the structures using modules that could be reused and, with some modifications, would allow the group to craft as many different buildings as needed.
“We originally thought about using the same nodal techniques as we used for the flying village in our asset manager, but that turned out to be impractical: It was too complex a structure for a single building,” explains Clarenc. “Instead, we created a library of parts, like rooftops, walls, and blocks, and assembled them by hand.”
The actual building destruction was done with a rigid-body dynamics system. “If you want to break a building into two pieces, you must also model the interior and break everything into small pieces,” explains Emilien Dessons, special effects supervisor. “We tried to use an algorithm-based shattering technique, but the shattering was equal and ugly, so we looked for a chaotic shatter with a graphically appealing shape. We manually modeled everything, shattering each piece with a reference drawing. We also added classically instanced particle systems to the destruction scenes to bring life to them, with a large amount of compositing and different smoke techniques added in as well.”
Furthermore, the crew used two modeling LODs, one with as few faces as possible, and another with beveled angles so the team could automatically subdivide the objects depending on the needs during rendering. “You could probably say we had a third level, where the objects were detailed according to the textures for extreme close-ups,” Clarenc says.
Hector the blue dragon proved difficult to model. He had to squash and stretch all parts of his body, including his eyes and teeth, without adversely affecting the skinning and blendshapes.
Devil in the Details
While atypical, the environments nevertheless had to look realistic, in contrast to the character models, in keeping with the style of the movie. “Guillaume Ivernel’s references were very detailed, and we had to translate that level of detail in every texture we created,” says Fabien Polack, texturing supervisor.
The environmental textures were fairly traditional, a combination of texture maps with simple shaders. “The challenge for us was more artistic than technical; we had to match Ivernel’s vision and enhance it whenever possible,” says Polack. Due to time limitations, the artists created ground shaders that worked for almost every shot—from expansive shots to close-ups. They also painted detailed textures but were careful to not fall into a common trap by “over-texturing” objects. “Like painters, we removed some material at the end of a texture to make it look more natural,” he adds.
Every set, Polack explains, was surrounded by a sphere mapped with a high-resolution matte-painted sky, provided by the matte department or Ivernel himself. Most of the time the reflections were real blurred reflections, and less often, environmental maps.
Every map was created, at minimum, at 4k resolution, sometimes 8k, and on occasion, 16k. The group provided the traditional set of textures—color, specular, bump, reflectivity, and, depending on the context, roughness, displacement (usually at 16 bits), and normal. Then, each high-definition shader was converted to mid- (2k) and low-def (1k) depending on the shots. The bump factors were also driven by a “distance-to-camera” node.
Despite using what Clarenc calls “a classic modeling setup,” the group still ran into some issues. “For instance, our renderer doesn’t have an angle-threshold limit to calculate the phong smoothing, which means that all our low-def models were totally out of shape and all the details were completely lost. So we developed a tool to convert the Maya hard edges—which are nothing but an edge-bound angle limit for the phong smoothing—to ‘real’ hard edges that were vertex-based with doubled adjacent edges to stop the phong from averaging values across them. And, voila, we had our low-def Maya objects in their crisp, detailed glory. While it doesn’t sound like a big deal, it soon became vital for the completion of the huge sets at the end of the movie.”
To render the images, Mac Guff had a 650-processor renderfarm available—small in comparison to Hollywood standards—though on average only 200 to 300 processors were used per shot. To keep the rendering times manageable, the crew was encouraged to keep each shot to a rendering time of less than two hours per frame, no matter the number of layers. Nevertheless, working in layers gave the artists more control over the final result.
Despite the nodal structure for the large scenes, an ever-growing library of modules and buildings in the asset manager, and a way to maintain the sharp angles of its low-def models during rendering, these solutions were not enough when it came to addressing the biggest problem of all: the world cemetery, where 200 shots occurred. Like the floating village, the world cemetery at the end of the movie was a single, huge set but contained dozens of detailed buildings, and the camera continuously moved through them.
“The animation was done, and there was no way we could change the layout, but the number of buildings was so large that none of our machines could render the scene,” explains Clarenc.
The team had to drastically reduce the polygon count; they did so by baking the high-definition models onto lower-definition models—sometimes as much as 10 times lighter. This involved both normal and color maps so that no one would be able to tell the difference if the objects were far enough from the camera.
The team used Mac Guff’s Symbor to generate the characters’ hair and the fur on their costumes. For each actor, they had at least three hair systems of varying densities and thicknesses.
However, that was the easy part. Afterward, the developers had to find a way to switch from the low-def models to the high-def versions, and vice versa, both in Maya and in the studio’s proprietary MGLR rendering software. This enabled the artists to arrange the layout for each shot and add details where necessary while keeping the “non-important” zones as light as possible.
“If our developers had not succeeded in doing so, the last sequence would have looked simple and empty,” says Clarenc.
VFX and More
According to Dessons, the crew used a significant amount of simulation throughout the movie. For instance, the team animated groups of bats and rabbits using a behavior-based system. In the few occasions where there was water, the artists used simple transparent and refractive shaders.
Mac Guff also developed new techniques to calculate clouds, reducing the overall time. This proved invaluable for a “cloud sea” sequence—in which the flying village is riding the cloud sea and dragons are crossing back and forth—containing a number of camera moves and set interactions. Here, the crew redeveloped the old-school ray-marching technique from the raytracer to get a real volumetric effect, although it resulted in huge rendering times (about 20 hours a frame).
This technique was also used for the dragon trails, enhanced with a 2D motion-blur process.
The raymarching technique also provided the opaque smoke throughout the film. “We also needed a smoke that reacted precisely with the environment. The ray-marching is a sphere-based technique whereby you build your smoke with spheres, so we couldn’t get that interaction,” says Dessons. So, the group used a voxel-based technique for a volumetric effect. Building on that technique, they generated a particle system that emitted voxels for a waterfall scene; this enabled the artists to control the water animation.
A Small World
The exciting and universal story line, the large personalities of the characters, and the stunning visuals make Dragon Hunters a movie for everyone. “This film is an example of why I am working in the animation business,” says Coldewey. “And it is a great example of how Europe works together. Think about it: The French and the Germans produce something in English for the international market. How cool is that?” Technically, the film utilizes the best techniques available, borrowing from US animation schools, French artwork, and Japanese image layouts and atmosphere. Will that be enough to satisfy US audiences? “That’s difficult to say,” says Coldewey. “To get a family feature to US theaters, you need to invest more than we had to make this movie. Yet, I think there is a big audience in the US asking for this type of story.”
Perhaps so. As Qwak points out, the movie is full of contrasts. The tone of the characters is cartoony, even though they’re textured like real ones. “We like the contrast between the adventure/suspense, and the sad/funny scenes,” he says. Emphasizing this contrast, Qwak describes Dragon Hunters as Tom and Jerry go to Sauron’s place (in The Lord of the Rings).
“Everything is blended, just as it is in real life,” Qwak adds.
Karen Moltenbrey is chief editor of Computer Graphics World.