is a story about a small but determined bunch with big dreams and a desire to prove themselves, and how, by using their willpower, courage, and “birdbrains,” they were able to beat the odds and triumph where better-qualified colleagues had failed.
On one level, that summary describes the story line for the new CGI feature film Valiant, about a wood pigeon who, despite his diminutive size, harbors a big ambition: to become a member of the Royal Homing Pigeon Service for England’s Royal Air Force during World War II. When given the chance to realize his dream, the pigeon overcomes seemingly insurmountable hurdles and, to everyone’s astonishment, succeeds in what his peers believed to be an impossible mission.
On another level, that description can be applied to Valiant’s filmmakers, who knew that their mission of creating a major studio-quality feature at half the typical price and in half the usual time was considered a lofty ambition. Yet, despite the huge risks and challenges, the group accomplished these ambitious goals.
“We did what many said was impossible, but through perseverance, dedication, and creative thinking, we were able to complete a $70 million film for $40 million,” says coproducer Curtis Augspurger, a seasoned visual effects artist (Shrek, Scooby-Doo, Batman Forever). “We also established a new bar for animated digital filmmaking that has been proven fairly successful in Europe but is only now being tried in the US.”
The concept for Valiant
, which originated with UK writer George Webster, eventually caught the attention of producer John H. Williams of
fame, who was looking for projects he could produce outside of the DreamWorks umbrella. Williams honed in on the coming-of-age story that’s told from the perspective of birds and beasts, and enlisted the help of Augspurger and Buckley Collum, also a coproducer on the film along with Eric M. Bennett. Altogether, they established a brand-new animation facility that was an offshoot of Williams’s live-action film studio, Vanguard Films.
Their goal in establishing Vanguard Animation was to break the “$1 million per minute of animation” barrier that’s the norm for high-quality CGI features in the US, by using a “non-studio” approach. “Although it had not been done before, there was no reason we couldn’t make a CG animated film of this caliber for half the usual price by thinking outside the animation studio box,” says Collum. In particular, Vanguard streamlined the infrastructure and applied a visual effects approach to the production. It also eliminated a tremendous amount of mid-level management. “The credits for Pixar’s The Incredibles
are enormous; in comparison, we had just 35 animators,” Collum points out. “Indeed, there are perks with a large organization, but we were efficient and could get things done fast. We knew from the beginning where we wanted to go with this film, and stayed on course throughout the production.”
Once the film was green-lighted in January 2003, Vanguard spent eight months of preproduction in Los Angeles, working on character and location design, character modeling, storyboarding, and animatic creation. In September of that year, the group shifted the physical production to a new facility in London for the production phase, which included set modeling, character animation, lighting, rendering, compositing, post, and editing.
Using an atypical approach to CG filmmaking, Vanguard Animation successfully completed its first mission: Valiant, about a wood pigeon and his flock of misfit friends who help the Royal Homing Pigeon Service.
While creating a new studio was fraught with challenges (see “Building a Birdhouse,” pg. 14), it also had its advantages. “We didn’t have to carry a legacy paradigm into our production,” says Collum. “Every new film tries to push the animation bar a little higher, and some studios are hampered with an existing infrastructure that includes older equipment and a pre-existing production paradigm into which a new production doesn’t fit well.”
Vanguard decided to use commercial packages in its pipeline and refine those tools as needed, rather than develop all its own software from scratch. To this end, the group based its framework on Alias’s Maya, mainly because of its robust tool sets and the ability to extend them via Mel scripts and available plug-ins. Another advantage was that most hires would be familiar with the content-creation program. In the end, Augspurger estimates that about 80 percent of Valiant
was done with off-the-shelf products, while the remainder was accomplished with in-house tools.
Originally, Gary Chapman assumed the role of the film’s character designer, but his ideas for the story, the settings, and the music made him an ideal choice as the film’s director. “One of my main concerns was establishing a look for the film. It’s a comedy-adventure, but I thought it was important to have some sort of homage to reality,” he notes. (Pigeons have saved thousands of lives during WW II, and 31 of the 53 top honors given to animal heroes have been awarded to pigeons.) “At no point did we approach this like a cartoon.”
As Williams points out, the team tried to keep a consistent period look in the overall design (objects, gear, backgrounds), but with some artistic touches that make it feel contemporary in its tone and subject. And the jokes and humor facilitate the story being told, as opposed to being based solely on pop culture a la the Justin Timberlake-Cameron Diaz reference in Shrek 2
. “We had to sell our jokes because they weren’t contemporary.”
In the high-flying CGI movie, the little pigeon Valiant (voiced by Scottish actor Ewan McGregor) and his misfit friends join the elite Royal Homing Pigeon Service (RHPS), which has suffered great losses at the claws of the lethal enemy falcons, commanded by the ruthless General Von Talon. While not fully prepared for duty, the pigeons-with the German falcons on their tails-are sent to retrieve a message from the French Resistance in occupied France and deliver it to the Allied Forces. When Valiant’s feathered friends find themselves in danger, the little bird not only saves his wingmen, but he also saves the day.
“The story is one that every child will be able to ascribe to and every adult has experienced: It is the plight of someone who is told they are incapable of doing something because they are too small or too young, and then they go off and prove themselves,” explains Augspurger. Valiant
, though, is told from a bird’s-eye viewpoint: pigeons, falcons, and other fowl play the lead roles, while white mice co-star as members of the Resistance. Humans appear infrequently, and when they do, they are obscured, so as to not pull the viewers, who are mainly children, out of this animal-centric world.
The artists used Maya’s Cloth to create the simulation for the black leather cape worn by the evil General Von Talon (center) in approximately 150 shots throughout the movie.
Yet, telling the tale from an animal’s perspective presented Vanguard with one of its greatest challenges: The team had to feather and fur every character-a task that was not in the group’s original flight plan. “We thought we would make photorealistic and visually appealing texture maps for the birds,” says Collum. “But in the early phases of preproduction, we challenged our effects gurus to devise a method of implementing feathers on the birds in a creative way that not only looked good in tests, but also could be used in production. So we shifted the concept from faux feathers to real feathers, and it paid off from a visual standpoint. But it was not without its share of headaches as we tried to improve the look.”
When the artists began the project, they were using Maya 4, which was limited in how it handled subdivision surfaces and polygons, particularly with the controls that would be needed for the surfacing. “But we knew we could take the surfaces we were generating in Maya and export them to [Side Effects Software’s] Houdini,” Collum says. “So our pipeline became more complex after the decision to do feathers and fur.”
Because of the tight production schedule, the group built a sophisticated character rig in Maya that was common to all of the birds and, with some adaptation (clipping the wings and adding a tail), was used for the mice, too. The rig contained a skeletal system under the face that defined the deformation and blended the morph targets that controlled the facial animation, particularly for achieving the phonemes and portraying emotion. It also allowed the end five feathers of the birds to be used as fingers as well as part of the wing.
According to Augspurger, the common rig system ensured that every animator would be familiar with any character’s controls. “As the production shifted, we could move our artists from characters that were temporarily out of production to others that were in production, thereby maintaining a constant work flow,” he explains.
Furthermore, the team created a customized plug-in called Chanko that enabled the animators to create a library of poses and clips-facial and body poses, performances, and so on-that were shared among all the characters, thereby giving the shots a consistent look. In addition, the coding team, led by Mat Selby and Manne Ohrstrom, set up referencing technologies and asset management tools that allowed the artists to access the key maps and positional maps, as well as the unique wardrobes of the characters, which could be swapped out on a per-scene basis at render time.
The images illustrate the feathering process for the character Von Talon: (from top, left to right) the initial guide feathers, the changed behavior for the guide feathers disturbed by the wardrobe, the RI behavior for final feather placement.
According to Rod McFall, character development and pipeline supervisor, each of the 30 main bird characters went through a rigorous grooming process that formed the basis for the feathering throughout the film. After modeling and rigging the characters in Maya, the team imported them into Houdini, where shaders were assigned. Guide hairs, grown from the surfaces, defined how the feathers looked and behaved-for instance, if they were ruffled or flat. Texture maps, meanwhile, dictated the type of feather, its directionality, color, and other properties that guided how they looked and moved within a given shot.
Each bird model had approximately 50,000 feathers, and once the bird got its feathers, it retained them throughout the film, since altering the number of feathers (because of their size) would have changed the visual look of the bird. Initially, the artists contemplated rendering the feathers and then generating an exact match for the displacement map, so they could render the displacement maps (and not the feathers) at a distance, while in the far background, they could simply use a flat color projection-all of which, in theory, would speed the rendering. However, this shortcut did not provide the desired timesavings. “The shaders and paradigm were clean, so it wasn’t as painful as we had initially anticipated,” says Collum of the rendering, which took between 5 and 20 minutes per frame.
Later, the feather information was exported as a Pixar RenderMan shader, lit within Maya, rendered in RenderMan, and composited into layers.
Now and again, however, the artists faced a recurring problem of pops in the feathers during the renders. Rather than redo and re-render the feathers in these shots, the group employed a 2D solution, using the Foundry’s MotionRepair tool within Furnace, a plug-in for Apple’s Shake, which the team used as its main compositor. “With the motion-estimation technology, we were able to interpolate areas from the surrounding frames so that we wouldn’t have to re-render the models and could simply replace the frames with artifacts by analyzing the surrounding frames,” Collum explains.
The team generated the fur for the mice in a similar way, using RI curves within Houdini, to which shaders were applied. In addition, the group used its own InfraFur plug-in to control the hairs so they would react properly when interacting with the tiny armbands, bandoliers, berets, and other items worn by each furred fighter.
Indeed, the feathers and fur presented a huge difficulty for the group; but having the characters’ wardrobe and military regalia interact with those surfaces was even more daunting. To achieve this, the group, while developing the grooming techniques, also defined the interaction between the feathers/fur and the clothing and objects within Houdini.
Using Houdini’s Attribute tool, the team extracted data from the clothing, defining areas where the feathers would need to push down and the gear would need to push up on the surface, thereby leaving a slight gap between the feathers and clothing.
Then, in instances when the feathers actually penetrated the gear, the animators used a simple compositing technique to render out the gear separately with a “holdout” map for the bird, and then used Shake to rotoscope little patches of the gear on top of the animated feathers. As a result, the team spent only an afternoon re-compositing a scene as opposed to an entire night re-rendering it in order to accommodate these intersections.
Because the Attribute Transfer tool only works procedurally based on proximity, the interaction animated as the relationship between the clothing and the bird’s surface changed over time. Then, once the studio networks were in place, the team set up a Web-based system that initiated the pipeline using whichever combination of character and clothing was needed for the shot. This ranged from pliable backpacks to hard helmets and medals to stiff leather straps.
Feathering the characters was challenging; having the feathers interact with the clothing was even more difficult. This task was accomplished with Houdini’s Attribute tool.
The process was put to the test in a sequence showing the pigeons dressing for their mission. As Augspurger points out, audiences usually don’t see many digital characters getting dressed in films because of the difficulty of the cloth interaction. “We took the technology as far as we could for the sequence-before we broke it, that is,” he adds.
Yet, the ultimate gear/feather challenge was presented by the uber bird, Von Talon, who sports a leather cape, created with Maya Cloth. The cape had to interact properly with the falcon’s feathers in approximately 150 shots.
“Unfortunately, we didn’t schedule time for the wardrobe, but it was something that the director wanted because it gave the characters a look and feel that was contemporary for their time,” says Augspurger, “and the authenticity of the wardrobe was key to selling that.” Therefore, nearly every shot contains wardrobe against feathers or fur.
After 18 months of production, Vanguard completed the movie-slightly ahead of schedule and budget. In all, the team created 1200 shots for the 75-minute film, thus proving that a featherweight production paradigm is stronger and more viable than many had previously believed.
“We just didn’t take no for an answer,” says Augspurger. “Also, we brought a visual effects background to the production process, which is slightly different from traditional animation in that instead of using the render tools-which are laborious and time-consuming-for fixing little problems, we took a different route and overworked our compositing team instead, saving both time and money.” (See “Foul Play,” pg. 16.)
As a result of this work flow, the team earned its wings, hitting 75 shots on average per week with a staff of only 150; in comparison, large studios with 600 artists usually accomplish 50 shots per week. “We achieved our goal and did so without compromising the aesthetic we were trying to achieve,” Augspurger notes. “We bit off more than we could have attempted a few years ago, but today’s tools allow you to envision things you couldn’t have even dreamed about five or 10 years ago.”
Vanguard used Houdini to fur the French Resistance mice. Because the hairs were so fine, they could be rendered at variable rates (100,000 to 1 million hairs), depending on the shot.
was released in US theaters August 19, though it debuted this past spring in the UK, where it soared in the top slot for a few weeks. Now that
has flown the coop, Vanguard is keeping its eye to the sky, waiting as Williams hatches the studio’s next project.
The fate of Vanguard Animation’s ambitious plans to create a studio-quality CG feature for half the price and in half the time as was the norm rested on the establishment of a streamlined pipeline built around commercial software. It also required a visual effects approach to CG animation that relied on compositing, rather than rendering, to make quick fixes.
After months of preproduction in the US, Vanguard transformed part of the former Ealing Studios in West London into a digital facility of the future, though it had to endure snags that occur when building a new facility, from getting the servers up and running properly to testing the tools and techniques during the actual production.
On the logistics end, to qualify for incentives offered by the UK, Vanguard had to employ a significant number of animators from the commonwealth and the European Union. As a result of this mandate and other factors, the multi-national crew-representing 17 countries and speaking 10 different languages-provided the team with a broad range of experiences that could be applied to the production, notes line producer Tom Jacomb.
On the technical side, the pipeline was built around Alias’s Maya, with Side Effects Software’s Houdini used for the feathering and the fur. Next Limit’s RealFlow, meanwhile, created the water and fluid simulations, and Pixar Animation Studios’ RenderMan accomplished the rendering. The renderfarm comprised 500 nodes of 1000 CPUs, including several IBM Blades; the remaining Boxx Technologies machines served as DCC workstations by day and renderers by night.
The team employed Next Limit’s RealFlow simulator for some fluid effects, though Maya was used for this shot.
“From the time we began budgeting the film and negotiated the hardware deal, we saw a threefold performance increase [in the hardware],” notes coproducer Curtis Augspurger. He points out that in a room a quarter of the size, Vanguard was able to accomplish the same rendering power for $2 million as PDI/DreamWorks did after spending $20 million when it made Antz
Another large savings resulted from Vanguard’s choice to use Apple’s Final Cut Pro for editing, rather than an Avid system, allowing for seven editors as opposed to two. Finally, for storing the digital assets during production, the group used a Network Compliance NetApp 940. Moreover, the team used a DI approach throughout the production, working solely with digital media on hard drives as opposed to film.
“To accomplish our goals, we stood on the shoulders of giants who have cheapened the price of technology and increased its power sevenfold,” says Augspurger.
Thanks to Vanguard’s carefully constructed pigeon coop, the studio was able achieve the goals it set with Valiant
. And in doing so, the group broadened the CG animation talent pool in London, and Europe for that matter, neither of which had been exposed to a digital feature of this scale before. And, the paradigm shift that
represents is likely to make major studios rethink the way they deliver animated films. -
To cut costs, the Vanguard Animation team had to avoid getting pigeonholed in their work flow approach, thus forcing the artists to look beyond the usual tried-and-true-albeit time-consuming-solutions often employed during CGI feature animation.
As a result, Gray Horsfield, CG supervisor, brought a good deal of transfer technology to the table, light baking being one of them. Prevalent in the gaming world, light baking for Valiant
was accomplished within Mental Images’ Mental Ray, and used mainly for the “colder” interior environments, such as the evil Von Talon’s lair. By light-baking the sets, the group could do an initial light pass with global illumination or radiosity lighting, then “bake” those maps into the color texture sets, from which individual renders were based.
By using this technique, the group no longer had to render the sets with multiple shadow-casting lights. Although the upfront costs were high, the subsequent savings for the individual renders were enormous, maintains coproducer Buckley Collum. “With this method, we saved a great deal of time on repeated renders, free from lighting and shadowing calculations involving many lights in complex scenes,” he says.
The team used light baking, a technique frequently employed in gaming, to achieve the lighting effects for the film’s indoor scenes.
According to coproducer Curtis Augspurger, a number of independent film studios are pursuing game development solutions for use in their productions. “The real inventions are coming from where the big money is being spent and made, and that is in the gaming industry,” he says. “There, they are pushing technology to be faster and better looking, and they have taken this to a new level in just the past three years or so. If you look at the trends being established by game developers, you’ll likely see the visual effects industry picking them up soon thereafter.” However, while game developers will hit a wall with these techniques in order to meet the real-time demands of their genre, the digital filmmakers can step in and use the technology for netting a higher degree of realism, he adds. - KM
may be limited in the diversity of the animal characters, but not so for the environments through which they fly. Nearly all the backdrops-from the tranquil English countryside and the bustling Trafalgar Square, to a busy air base and war-torn France-are 3D, augmented at times by mattes and set extensions. “We’re all over the map with our environments, but they keep pace with the emotional pulse of the film,” states coproducer Curtis Augspurger.
One of the more popular backdrops in the film is a sky populated with a range of clouds created using a 3D volumetric cloud renderer based on a program by Joshua Schpok, a researcher from Purdue University. Created as a stand-alone OpenGL-based program, the cloud simulator was retooled by the researcher to fit within Vanguard’s Maya-based pipeline, allowing for the real-time generation of realistic types of clouds, from ominous dark formations to those of a thin, wispy variety. - KM
The artists crafted a wide range of backgrounds, from expansive shots of the French countryside (top) to the busy streets of London (middle) to the wide, open sky (below). For the clouds in these shots, the group used a real-time volumetric renderer.
Karen Moltenbrey is an executive editor at Computer Graphics World