Big Buzz
Issue: Volume: 30 Issue: 10 (Oct. 2007)

Big Buzz

Dreamworks tackles scope and scale to infuse bee movie with Jerry Seinfeld's humor
By Barbara Robertson
 
A short while before DreamWorks released its animated feature Bee Movie, writer and producer Jerry Seinfeld fielded a comment at a screening from an audience member who had been reading about a bee crisis in the news.

With perfect timing, the immensely popular master of wry humor re­sponded, “The crisis in the bee community right now, the real bee community, that colonies are collapsing, is actually a publicity stunt by DreamWorks to sell the movie. As soon as it’s released, we will replace all the bees.”

In the film, Seinfeld’s humor takes wing, literally, in the form of Barry B. Benson, a honeybee. When we meet Barry, a recent college graduate, he is resisting joining the workforce that makes New Hive City’s sole product: honey. Instead, he buzzes outside the hive, flies over Central Park in New York City, and tumbles into the apartment of a beautiful and quirky young florist named Nancy (Renee Zellweger). When she saves his life, he can’t help but break the bee world’s cardinal rule: Never talk to humans. Despite their differences in size, the two develop a friendship, and Barry begins learning about the human world.

Barry: “TiVo. You mean you can just freeze live TV? That’s insane.”

Vanessa: “What, you don’t have anything like that?”

Barry: “We have ‘HiVo,’ but it’s a disease. It’s a horrible, horrible disease.”

At the screening, before Seinfeld showed the film, an audience member asked, “Is this movie about something?”

“No, it is not,” says Seinfeld. “It is more of the same nonsense and drivel that I have been foisting on the American public for years, and doing very well with it I might add.”

The film, of course, is about something. It’s about Barry’s awakening sense of himself, his discovery that humans steal the bees’ honey, the lawsuit he instigates, and its consequences. And it’s about the relationship between Barry and Vanessa, between two characters from different worlds.

When directors Steve Hickner and Simon J. Smith recorded dialog for the film, they recorded Seinfeld and Zellweger together. “It made the conversational parts sound so natural,” says Smith.

Preserving and enhancing those and other comedic performances challenged the directors, the animators, and the production crew. “This movie is quite fragile,” says Smith. “With comedy, you can easily break things. You have to make every joke work.”
 
Big and Little
Scale was arguably the biggest impediment. Barry and his hive-mates are less than an inch tall (.825 inches); Vanessa and the other humans are human size. “When you have two characters talking to each other and developing a relationship, the bread-and-butter shot is a ‘mid two,’” says Smith, “a waist up shot of the two characters.” That is not possible when one character is a bee. It was even difficult to cut from one character to the other.

“We learned that we couldn’t go below a 28mm lens on Vanessa, or Barry would look distorted,” Smith says. “We didn’t cheat by using one [virtual] lens for Vanessa and another for Barry in the same shot.”

While the directors and layout artists solved these creative issues, the technical crew tackled another problem: To fly Barry across New York City, the software had to handle small size and vast distance simultaneously. “We had to retool our entire software tool set to deal with this,” says Doug Cooper, visual effects supervisor. “If a bee is on the rim of a coffee cup, everything is within a few units; everything is small. But, when he flies down the side of a building and along a street in New York, and the camera stays close to him, we need precision to the hundreds of thousands of an inch while moving him millions of units down the street. That’s where single precision breaks down.”
 


          

DreamWorks relies largely on custom software, with an assist from Autodesk’s Maya. “Our R&D team rewrote every piece of software to have double-precision floating point,” Cooper says, “and we worked with Autodesk’s custom engineering team to fix similar problems with Maya. Once the R&D teams solved this problem, it literally became a non-issue.”

Sliding Scale
But, that wasn’t the only issue with scale. To help the layout artists more easily compose shots, the character modelers and riggers made all the bee characters scalable. Animators could change the bees’ sizes dynamically from shot to shot and within a shot.

Cooper provides this example: “We had a case in which Barry was on a coffee cup in the foreground. It’s a nice close-up of him with Vanessa—far enough behind [so] that the shot looks good. In the scene, he flies off the cup, past Vanessa, and lands on a cereal box. By the time he gets there, he disappears. That’s the real perspective of the scene, but it doesn’t work for storytelling.”

To make Barry and other bees visible when they fly away from camera, the riggers gave animators sliders that dynamically changed bee sizes throughout a scene. And, not just their size: When animators scaled the bees, every piece in the pipeline looked at the scale and tweaked itself automatically. Fur grew longer and thicker as needed, displacement and bump maps adjusted, shadows changed size and sharpness, and so forth.

So that the artists could see whether all the pieces scaled properly, and to validate surfacing and lighting, the technical crew developed a new turntable technique. Rather than trying to judge a bee as it would appear in the shots, from a tiny spot in the distance to a large creature in a close-up, the turntable automatically moved the camera closer or farther away so that the bee maintained the same size in the frame.

“It looked like a straight spin,” says Cooper. “But the character was actually shrinking and scaling, and we could see immediately whether that was working or not.” An example of not working: “One time, the hair grew into a giant ’fro because it wasn’t scaling with the character.”
That kind of scaling was necessary only when the bees ventured into the human world. In their own hive, they were six feet tall. “We built the bees and Hive City to human scale so that we could shoot the scenes with a normal camera and use normal language to discuss things,” says Cooper. “That way, we could say that the molding around a door should be a couple inches thick. If we had worked in bee scale, our minds would have exploded trying to figure out the relationships.”

Scoping It Out
All told, the crew created 39 locations within the human world and the bee world. “Very early on in this movie, Jerry [Seinfeld] made it clear that he wanted the audience to see the world the characters live in, not just the stage,” says Cooper. “We had 43 sequences and almost as many locations, and the locations had sub-locations within them. To say Central Park is a location is like saying the earth is a location. We had a tremendous amount of stuff to build, surface, and light.”

Modelers and effects artists constructed the human world using straight lines and formal angles, putting approximately 150 buildings on the streets of New York City and Pasadena (think Rose Bowl Parade), adding cars and streetside accoutrements, and decorating the building interiors with hundreds of props. In addition, they filled Central Park, designed from a bee’s point of view, with trees and brightly colored flowers.
 
 
The bees’ world existed inside a three-walled hive built with hexagons, wax, and honey. The outside wall led to the outside world. A space between that wall and a middle wall contained the hive’s suburban streets and houses. A balcony extending from the “back” of each house was built into the third wall, an interior wall that overlooked the inside of the hive, where busy bees produced honey in New Hive City’s Honex factory. Creating all these worlds required approximately 3000 models plus matte paintings.
 


When you’re inside the city looking up at the interior wall with the balconies, in most shots, you’re looking at matte paintings. “They’re 2.5D paintings projected onto geometry,” explains Cooper. “But, we had CG bees hanging out on the balconies, so we had to project and register the paintings perfectly to have the bees in the proper places.”

Concept Models
To help speed production for this film, the crew forged a tighter link between concept design, modeling, and surfacing than with previous films. “We had a modeler move into the art department and work on designs in 3D,” says Cooper. “And a surfacer spent several weeks in the art department doing clothing and texture design.”

Sometimes the artists created environments in Maya, took snapshots of those environments, and then created 2D artwork from those models. Sometimes the artists continued working in 3D, which helped them communicate their vision, especially for the complex beehive in­terior. And sometimes they created designs so detailed the modelers could tidy them up and install the designs as final models.

“We received art in a number of ways,” says Ardie Johnson, head of modeling. “But, the key was that we were working with the art department. At peak production, we would spend seven to 10 hours a week working together, hands on, with character designers, the production designer, and the art directors. It was fantastic. We grew artistically as a modeling department.”

In character design, for example, the modelers learned that they worked better from 2D art, from one “approved money pose” rather than from clay maquettes or from orthographic side, front, top, back, and bottom drawings. “We said, ‘Give us the moment with emotion and we’ll model something,’” says Johnson.
 


A drawing of Barry, for example, captured the emotion of the moment when he announced that he was going to sue the human race. “He’s in front of a lot of mikes,” says Johnson. “I worked with the art directors and modeled a quick pose of Barry in that moment. Jerry [Seinfeld] came in a week later and he loved it.”

Because people like to hold a model of a character in their hands, touch it, and rotate it, once the modelers had polished the 3D models, sculptors created clay maquettes from the computer models.

Worlds of Fun
Knowing they’d need to fill the locations with hundreds of props, the crew also devised new techniques that they call “contextual,” or “holistic,” modeling to make the task more efficient than on previous films. In the past, modelers would build a long list of props for a scene, give those individual elements to the layout artists who showed them to the director, tweak the props as needed to please the director, and then the layout artists would dress the sets. To fill Vanessa’s apartment with furniture and other items, for example, they would have carefully built each object and sent it through the approval process individually.

This time, they quickly generated models in a rough form, put them all together in the set, and looked at the entire scene with production designers and art directors before taking the models further. “We put our modeling money where the camera was,” says Johnson. “We never had the directors look at a single object by itself unless there was a particular reason to justify that.”

As a result, rather than spending time on props that would never be bigger than 15 pixels wide, the modeling team focused its efforts on the hero props. And that, Cooper believes, not only saved time, it improved the look. “The design was more cohesive,” he says. “We didn’t have gorgeous hero pieces all competing for attention.”

That said, the modelers still needed to provide individual pieces to other departments for surfacing and textures, and those artists wanted the models oriented around a central x, y, z axis so they could put them on turntables. A proprietary tool called set buster made it possible.
“Set buster allowed us to export models from Maya and still preserve their placement in the set,” explains Johnson. “Once the models were approved, we gave the layout artists the position of the lamp on the table and the chair in the apartment on the eighth floor.”

Stylism
Smith uses the word “stylism” to describe the design style, the look of the environments and characters. The story drove the design.
“If you have a relationship between a bee and a human, you can’t have them look realistic because that would just be icky,” Smith says. “So, it was tricky getting it right. We painted everything stylistically. But we lit it in a realistic fashion.”

Vanessa was the most difficult for the artists, according to Cooper. “We finally hit on a Photoshop painting that everyone gelled around which had the right quality for her skin in terms of specular and texture,” he says. “We made sure her textures came from sources that weren’t photos of skin.” They used, for example, electron scans of aluminum pieces and other images with organic patterns to create skin textures in a non-photorealistic style. For her shoulder-length hair, the team borrowed technology developed for Shrek the Third (see “Merry Tales,” April, pg. 12). 

The bees were simpler, except for their wings and their color. On previous films, artists had used a volumetric approach to simulate the look of fairy wings. But, bees beat their wings thousands of times a second, and the volumetric approach would have been too expensive. Fortunately, Bee Movie’s stylized look worked to their advantage.
 


“We took inspiration from dry-brush animation,” says Cooper. “We scattered particles across the wings, and then swept through the arc of the wing and motion-blurred those particles. That gave us a streaky quality rather than a spread-out fan of wing blur.” To heighten the look, they also added specular highlights and glints inside the wings.

Having a method for quickly creating wing blur by using the studio’s particle renderer was particularly important for shots with thousands of bees. In one scene, for example, approximately 750,000 bees land a jet with help from another 350,000 bees on the tarmac creating flower patterns to direct the landing.

The crowd animation team used a typical rule-based crowd system for small swarms and bees walking in the city. For swarms in the thousands, they used a particle approach, rendering the bees in the studio’s proprietary particle renderer.

As for color, although the bees had a more cartoony style than the human characters and, thus, simpler textures, their color was finicky. “At the right hue, Barry was sunny, golden, happy to look at,” says Cooper. “A shade redder and he looked sunburned. A shade greener, you wanted to throw up. We walked a fine line getting the yellow balanced in every scene and every environment.”

The beehive, built with beeswax and honey (see “Honey, Honey,” pg. 21, and “Waxing Rhapsodic,” pg. 22), caused similar problems. If everything was golden, there was no contrast. “We had lots of warm, yellowish golden surfaces, but the time of day often dictated cooler light,” says Mike Necci, one of six lighting and CG supervisors. “Blue lights on yellow surfaces turn the surface green.” A special shader let the lighters use what they called “Tint” lights, with which they could control the surface color on a light-by-light basis.

For cloth, the character effects artists used Syflex to handle costumes for the humans and the bees. In one scene, a bee stings the lawyer for the humans during Barry’s court case, and the lawyer rolls around on the floor. “It was hugely difficult,” says Cooper. “We had to simulate a loose suit jacket with a character on top swinging around. We accomplished it by simulating two or three versions and blending them by hand.”

In another shot, “pollen jocks” bring Barry a jacket and help him slip his arms into the sleeves, put it on, and zip it up. “You usually don’t see a character in an animated film put on a piece of clothing because it’s really difficult,” says Cooper.

In Bloom
The pollen jocks also contributed to one of the most difficult shots in the film: the pollination of Central Park. Modelers created the basic shapes of the trees in the park, and then effects artists, using the studio’s procedural modeling tool kit, sprouted branches, leaves, and blossoms within those shapes. “Once we developed the rules for the look we wanted on one tree, it was easy to transfer the system onto trees,” says Allen Ruilova, lead effects artist. “The big thing was getting the growth to fit within the shapes and still look realistic.”
As the sequence begins, the camera is on a bicycle path. The trees are bare. Then, the pollen jocks fly into the scene spraying pollen from their backpacks like crop dusters. The camera pulls out and rockets into the sky to reveal a complete view of the park as it comes alive and blossoms.

It’s an apt metaphor for the production, as well.

“That shot taxed everyone, from modeling and layout to lighting and effects,” says Cooper. “But the movie is full of shots like that. The scope and scale of the film was a massive challenge, and a whole lot of fun. It’s satisfying to work on something for three years and then see it bloom.”  

Honey, Honey
 
We discover the Honex factory early in the film when we follow Barry as he learns about his job possibilities. In the factory, the bees operate such contraptions as a centrifuge and distillery, pull the honey like taffy, and scrape honey from honeycombs. The honey runs through factory tubes into tanks. It even fills swimming pools on the bees’ balconies. Effects artists generated the honey’s motion; the lighting department applied the shaders.
 


“The directors wanted to give a sense that we see the honey in various states of production,” says Allen Ruilova, lead effects artist. “We needed to create motion in different levels of viscosity, from runny nectar to honey.”

DreamWorks’ proprietary fluid simulator, FLU, allowed the artists to adjust the viscosity of the honey, but it didn’t give honey the elasticity the artists needed. “Imagine that you’re pouring honey onto a flat surface,” Ruilova says. “It folds. The folds form, they hold for a second or two, and then they melt.”

To create those folds, the fluid had to retain its physical shape even though the fluid behind it pressed on it. When the pressure became too much, it had to release and flow. To add that elasticity to the fluid simulator, the crew relied on Adam Bartgeil, then at UC Berkeley, now a post-doctoral fellow at Carnegie-Mellon University. Ruilova calls the result “visco-elasticity.”

“If you keep making a viscous fluid more and more viscous, it just looks slower,” Ruilova explains. “Visco-elasticity allows for a play between viscosity and elasticity. They work in tandem.” The new technology was particularly useful when effects artists scraped honey off the apiary’s honeycomb walls.

For other shots, the artists used bubbles to create the illusion that honey moved through tubes and in tanks and swimming pools. To move the bubbles in tubes, Ruilova used a wire deformer and shaped a particle simulation. “The bubbles flowed along that shape,” he says. Rather than use a fluid simulator for the swimming pool, Ruilova created the honey using a flat surface deformed with noise and procedurally generated ripples. For shots of Barry swimming in honey, he surrounded Barry with a force field that pushed and pulled particles to drive bubbles that reacted to the bee’s motion.

To sell the look of honey, the artists used a shading system to change the color of honey based on its distance from a light or the camera. “The honey in jars at the supermarket looks richer and thicker toward the center more so than near the edges,” Ruilova says. “And the honey looks thinner on the steps going into the swimming pool than at the deep end.”

Sweet.
 
Waxing Rhapsodic
The effects artists also worked with new shaders to give the beehive a magical translucent quality. “You can attenuate and color-adjust as light travels through the volume,” says Mike McNeill, lighting and CG supervisor. Although subsurface scattering might have accomplished the same effect, it would have been burdensome to use that technique on such a large scale. The designers created everything in the hive with wax; it all needed a backlit translucent feel.

“The technique is based on generating depth map-based calculations from the light’s point of view,” says Mike Necci, lighting and CG supervisor. “That depth map is used to calculate falloff and the color of light as it passes through the surface. We had separate lights for transmission and for the surface; we could change how transmissive the surface was on a per-light basis. Doing this on a per-light basis gave us tremendous control over the subsurface glow and the color of light on the surface.”
 

Barbara Robertson is an award-winning writer and a contributing editor for Computer Graphics World. She can be reached at BarbaraRR@comcast.net.