Viewing the situation from their perspective, displaced animals clash with suburbanites in Over the Hedge
By Martin McEachern
One hundred and eleven years after the Lumiere brothers first introduced the world to the art of motion pictures, film goers have grown accustom to a cinematic language forged in celluloid. The chemistry of film stock, properties of lenses, and the movements of lumbering cranes and dollies have combined to create a visual vocabulary that is readily familiar to audiences worldwide. While that language may seem archaic in a digital production environment—where exposures are flawless, shadows perfectly gradated, and virtual cameras fly through the air at the click of a mouse—PDI/DreamWorks’ fifth animated feature harnesses its power in away never before seen in a CG film. The result is a veritable looking glass through the eyes of woodland animals displaced by suburban sprawl.
Over the Hedge opens at the dawn of spring, as the animals awake from hibernation to discover a giant hedge has cropped up through their home. “It never ends!” yells a hyperactive squirrel, standing small before the massive wall of shrubbery that stretches endlessly in either direction. On the other side, their forest has been demolished, replaced by the newly-minted subdivisions of El Rancho Camelot Estates, where peculiar creatures called humans speak on cell phones, live in air-conditioned comfort, and plow over everything in their path behind the wheel of their giant SUVs—including furry little quadrupeds.
Fearing the new world, Verne, a fatherly tortoise voiced by Gary Shandling, manages to keep his family of mixed species on the safe side of the hedge—that is, until their loyalty is challenged by the arrival of RJ, a con-artist raccoon voiced by Bruce Willis, who lures them to the other side with visions of trash cans filled with nacho cheese corn chips, soda pop, and previously unknown treats. Little do they know, however, that RJ is only using them to steal a wagon load of human goodies he must repay to a bear named Vincent before he eats him for lunch.
The first to give in to temptation is Hammy (Steve Carell), the childlike squirrel with ADD and the speed of the Road Runner who wrestles with insecurities over his intelligence. The rest of Verne’s family soon defects, including Penny and Lou (Catherine O’Hara and Eugene Levy), a porcupine couple with a brood of little porcupines; Stella (WandaSykes), a skunk whose sassy attitude masks a deep fear of rejection; William Shatner’s Ozzie, a possum who plays dead with Shakespearean aplomb; and his much-embarrassed teenage daughter, Heather (Avril Lavigne), who eventually learns to value her father’s talent and embrace her possum identity.
RJ coaches Hammy to act like a rabid squirrel, while the directors use
long lenses and shallow depth of field to magnify the scale of the world
to forge greater intimacy with the viewers.
All images ©DreamWorks Animation, L.L.C. and DreamWorks, L.L.C
As RJ leads his unsuspecting friends into danger across the manicured lawns of suburbia, the group must contend with Gladys, the short-tempered president of the local homeowners’ association, and the Verminator, whom she hires to exterminate them. In winning his friends’ trust, however, RJ discovers his own need for family and is torn between his self-serving ways and his new found allies.
“The very first marching order I discussed with the crew was, ‘This is not a fi lm about how small the animals are; this is a film about how big the world is,’” saysco-director Tim Johnson. “I kept insisting that people consider the animals to be 6feet tall and the world to be a 150 feet tall.” To create that dizzying sense of scale, Johnson and co-director Karey Kirkpatrick experimented with a variety of techniques from traditional cinematography, including using wide-angle lenses.
“We thought at the outset we’d stick with the wide lenses, but sometimes that did just the opposite: It looked like a wide lens, so you immediately cast a judgment about the scale. On the other hand, we found that slightly telephoto lenses dida much better job of making the world seem bigger than it is,” says Johnson. “By throwing focus and isolating characters against their backgrounds, we ended up with a way to not only identify with the characters—because they’re the only thing in focus on screen—but to capture the overwhelming size of the world.”
The use of rack-focusing and shallow depth of field was horrifying to many, especially the lighters and texture artists whose work would be deliberately discarded. After all, it’s an anomaly to subtract rather than to add in digital animation. “In conventional photography, you begin with the whole world, then subtract what you don’t want to shoot, then subtract exposure, because you can only see fi ve stops on a photograph; then you subtract focus, because you can’t hold everything in focus. In CG, you usually start with a blank canvas and then add stuff to it, and you tend to over-add, I think. Making it a rule that we were going to throw out information allowed us to create a far more visceral film, one told from the animal’s point of view."
A Squirrel’s-eye View
Point of view (POV) was so important to Johnson, in fact, that prior to production, he screened Steven Zaillian’s Searching for Bobby Fischer for the animators, showing them how camera angles, shallow depth of field, and editing could be used to force the audience to see through the eyes of a child chess prodigy. “The movie is about a small boy, so the camera is always low,” says head of effects Mahesh Ramasubramanian. “Also, kids tend to focus on a few things; they can’t just observe everything, so there's lots of shallow depth of field. Again, if you do it right, people won’t notice, but if you overuse it, it can become distracting.” In the spirit of that film, Johnson attempted to never cut to a human’s POV or cut above the animals’ shoulder. “Again, if you treat them like they’re 6 feet tall, how often do you really put the camera over a person’s head in live action? Pretty rarely,” he says. “You usually keep the camera at tripod height, so we made a hard rule about that.”
From the 100,000-plus hairs adorning each animal, the surfacing department selected 500
to 2000 to be guide hairs, which the animators then manipulated with magnets to create
the disheveled fur of Hammy.
To further enhance the cinematic feel of the film, head of layout Damon O’Beirne limited the focal lengths used on the film to a set of Panavision prime lenses, including 10mm, 14.5mm,17.5mm, 21mm, and 27mm. Longer lenses, such as 40mm,50mm, and 75mm, produced the shallow depth of field. “When the background is out of focus, you will be drawn in much closer to the animal world and see the human world as ‘them,’ without having to think about it,” says character TD Nico Scapel. Since depth of field was so important, the R&D crew developed a way to visualize the depth of fi eld in previz shots. They also studied the naturalistic cinematography of Haskell Wexler, Conrad Hall, John Seale, and Caleb Deschanel, embracing lens flares, under and over exposure, and fully blackened shadows to make the film more experiential, especially in depicting the clash between the world of the forest and the “new world” of suburbia.
“This is like the Native Americans meeting the first European explorers, so we wanted to express that difference through their environment,” says Johnson. “The forest is very overexposed, with light cumulus clouds strewn across the sky. We also tried to use forgiving shadows, much more fill, and achieve a dappled look in the color-rich world. When we go to suburbia, we thought the huge stretches of asphalt and concrete would feel like the desert to the animals, sowe lit it hot and made the sky cloudless, save for the jet-engine contrails.” Johnson also chose colors you wouldn’t see in nature, like the cool pastels of the colored homes and the bright metallic of the cars. The directors also found the perfect way to put the animators in the mind-set of the animals. By having them stand next to the looming 70-foot-tallivy-laden wall of the DreamWorks Animation parking garage, they could imagine what the hedgewould look like to the animals. “It was mandatory that everyone in the crew spend a little time in the paws of the animals by going out and looking at this gigantic wall and imagining it had just shown up,”says Johnson. “It was a fundamental salute that we did to POV throughout the picture.”
Animal Animation
Because the entire film unfolds from the animals’ POV, the human characters often seemed distorted through extreme perspectives, forcing the artists to make slight alterations to their models. They also had to be pay close attention to the deformation of their shoes, which were subject to close scrutiny from the ground-level POV.
As the animals invade suburbia, the Verminator is charged with hunting
them down. Cloth animation for the humans was done with Syflex.
In crafting the performances for each character, Johnson points to three driving forces: the story line, arc, and emotional state of a character; the species and physical makeup of a character;and the movements inspired by the vocal performances. “So, when you’ve got a nervous character, like Eugene Levy’s father, coupled with his Midwestern, ‘golly-gee bubbly-ness,’ then you've got a slow, squash-and-stretch little bear whose movements suggest a water balloon-like motion,” he says.
To capture the springy, wobbling motion of the porcupines’ bellies, artists set up a dozen mass-spring particles over the belly area, then attached its ENET polygon cage (representing muscles) to these locators. This mass-spring system, coupled with the dynamic fur animation, proved far more effective than using soft-body dynamics in conveying a realistic feeling of weight and inertia. Animators could do a simulation pass and then change the amplitude of the simulation without having to run the simulation again.
All of DreamWorks’ character dynamics systems are pose based, which means they follow the animators’ poses, adding layers of follow-through and overlapping action that can be blended on and off during the shot. “This proved extremely useful in extreme action sequences, where the dynamics blend was often set to 10 percent or less,” says Scapel.
The Animals’ New Fur
Of course, nothing had a greater impact on modeling, animation, and work flow than fur. From the short pelt of RJ and Hammy, to the glossy black coat of Stella and the long quills of the porcupines, fur interactions became such a challenge that the character effects department imposed a “hug quotient” on the animators. “It’s hard to have furry characters hug, or even push through a hedge—through dense leaves and branches—without having intersection problems,” says Johnson. Complicating matters, some sequences required the characters to directly manipulate each other’s hair, as when RJ dishevels Hammy’s hair to make him appear rabid.
Because of the extensive use of shallow depth of field, global illumination
on all the characters, and Hierarchial Volume Data for self-shadowing on fur
and foliage, DreamWorks pooled the rendering farm of its two locales to
handle the demand.
To handle these challenges, Over the Hedge uses a new fur system that is an off shoot of the wig system used on Shrek 2 and Madagascar, and is the first from DreamWorks to offer full-body dynamic fur, meaning the fur can be dynamically animated from head to toe. The surfacing department created the fur for each animal using fl at strips of geometry that were textured to appear like a tube. For example, RJ has 1,656,680 hairs on his body, while Ozzie has 477,037 hairs, compared to his daughter, Heather, who has 403,649. The possums’ fur has two lengths—a shorter layer underneath and long hairs on top—and was difficult to control in the neck area, where the hairs would penetrate themselves every time the possums looked up. The porcupines have approximately 400,000 hairs and 400 long quills.
From the hundreds of thousands of hairs, the surfacing department selected 500 to 2000 to act as guide hairs, which the animators manipulated to create performance-based animation, such as with the tufts of hair around Stella’s cheeks. Like the wig system for Shrek 2, animators pulled groups of guide hairs with magnets. They also applied force vectors for wind and inertia to the magnets to simulate dynamic motion, creating follow-through and overlap on the fur. “The dynamics were done with the magnets themselves,”says Ring. “The magnets would bounce up and down as the characters walked, almost as if the magnets were suspended on a spring. That’s how the fur bounced. And, the animators could also override the movements of those magnets if they needed.”
Force vectors for wind and inertia made the fur bounce realistically.
But when Stella, the skunk, was given a feline makeover, her fur
was so shiny that the dynamic motion had to be greatly reduced.
Once the animation was completed, the character effects department checked for fur inter penetrations and solved them with a set of procedural tools called Smoosh. By casting rays off individual hairs, Smoosh could tell if they were intersecting anything. If a character’s hand intersected with another’s fur, for example, the artists placed a localized wind creator in their hand to push the hair away. For intersections that could not be solved procedurally, the artists loaded the guide hairs into Autodesk Media and Entertainment’s Maya, clustered CVs, bent, or flatten them down, and then re-imported the new key frames back into EMO, the company’s proprietary animation software.
Because fur added volume to a character, the animators had to constantly monitor the silhouette of each animal after it was rendered, to make sure the performances were preserved. In addition, the fur hid the signature wrinkles that made the facial expressions readable. “One thing that we noticed right away was that fur had a tendency to smooth out the animation, so the directors asked the animators to push the animation an extra 20 percent to compensate,” says Scapel.
To give a greater tactile feel to the fur, the artists used aggressive rim lighting on all the characters—a technique borrowed from the Dr. Doolittle films. By placing spotlights with controllable barn doors behind each animal, their coat of fur would glow diaphanously in the backlight, giving greater definition to individual strands as they swayed dynamically in the wind.
Depth of Character
While shallow depth of field can do many things—direct the audience’s attention, narrow a character’s focus, isolate them indislocated space, and add disorienting scope to the world—the blurred image is, ironically, more expensive to render. Shallow depth of field also impacted the lighting of the scenes. That’s because when images go out of focus, light sources tend to lose their brightness, forming cloudy circles of confusion.
“We did some experimentation with HDRI to keep highlights bright. So if you get a specular highlight as it goes out of focus, instead of going muddy gray, it remains white because it’s many times brighter than white,” says visual effects supervisor CraigRing. “But instead of using HDRI, we found we had more control if we rendered and blurred highlights as a separate pass, then controlled their brightness in compositing.”
Aside from the out-of-focus backgrounds, another major rendering burden was the grass carpeting in most of the environments. While crucial to embedding characters in a dynamic, living world, grass is extremely slow to render. “That’s because there's so much geometry and so many characters interacting with it,”says Ramasubramanian. To cope with the increased rendering demands, DreamWorks pooled the rendering farms of its Redwood City and Glendale studios. With approximately 3000 processors between them, the average render time per frame at film resolution was approximately 12 hours.
The grass, itself, was grown using the same system as the fur, and also employed Smoosh for animation. While guide hairs were rarely used for its animation, the grass shudders and undulates gracefully in the wind using force vectors applied at render time. Artists also applied wind vectors to a character’s feet to bend the grass under their footsteps.
Unlike the grass, the hedge is composed of a half dozenhand-modeled leaves in NUBS, its proprietary Non-Uniform B-Splines format, that were procedurally duplicated and scattered across the body of the hedge. Subsurface scattering shaders were applied to the outer layers of the foliage, accentuating the translucency of the leaves as light passes through, especially in back lit, overexposed shots, where lens fl ares bloom across the screen. Subsurface scattering can also be seen on all human skin, as well as the tongues, hands, ears, lips, and mouths of the animals.
The film’s directors embraced naturalistic photography, using lens flares
and under- and overexposure to make the movie more experimental.
Along with subsurface scattering, the company was able to fully harness a variety of lighting technologies under development since Shrek, giving a new standard to the company’s renowned soft, pastel look. The first technique was a consistent use of single-bounce global illumination on all the characters. This means that when characters walk over green grass, they’ll receive green bounce lighting. The bounce lighting from the ground also gave the shadowed regions a soft fall off. More important, however, was the prodigious use of Hierarchial Volume Data (HVD) to produce self-shadowing on fur and foliage. Normally, self-shadowing on densely arrayed pieces of geometry is accomplished using depth-map shadows.
But the aggressive application of HVDs, subsurface scattering, and global illumination did not come without a price. The lighting often became so busy and visually complicated that, without a way of simplifying it, audiences would have found it distracting. To solve the problem, the crew developed a new technology called Averaged Normals. When a scene was lit, there sultant lighting was reduced to 50 percent importance. The same scene was then relit using only the basic, underlying shape of the objects. For instance, a tree would be relit using only simple spheres representing the clumps of foliage. This lighting was then combined with the original lighting of the detailed tree to soften and even the shadows while still retaining the geometric detail. (DreamWorks has submitted its work with Averaged Normals, along with Sonic Hammy, as sketches for this year’s SIGGRAPH.)
The Leading Hedge
To create more cinematic camera moves, the team modeled and rigged dollies and cranes, then attached the virtual camera so it would move like its real-world counterpart. The final step in emulating the natural effects of light on film was to create and apply a film-response curve during the rendering phase. When the film-response curve is applied, it simulates film’s sensitivity to light; images will under- or overexpose more naturally, or fall off into deep blacks.
Reflecting on the legacy of Over the Hedge, Scapel sees the film as a coming-of-age for computer animation. Having assimilated the lessons of both live-action films and traditional animation, it is beginning to forge its own identity, he says, blossoming into a medium where each character can have its own animation language, and where wildly cartoony animations can coexist with the subtlest of emotional expression.
For director Tim Johnson, computer animation holds the capacity for “photo-unrealism,” and it is precisely this coexistence of the real and the surreal that is crucial to the success of the medium. “I have the theory that 90 percent realistic is fun to watch; 98 percent is not fun to watch. There’s a place where it becomes so life like that the tiny differences become irritating, if not a bit creepy,” he says. “While you don’t want your stylizing to be so off-putting that the audience is not bonding with the characters, you’ve also got to use the medium to express something imaginative. Merely imitating exactly what you’d get if you pointed a camera [at it] is a cop-out.”
Martin McEachern, an award-winning journalist and contributingeditor for Computer Graphics World, can be reached atmartin@globility.com