[ continued from page 1 ]
Another 80 shots were parceled out to Visionart, which was concurrently working on Godzilla. Unlike POP's shots, the lion's share of Visionart's talking-animal work was 3-D, involving everything from muzzle replacement to full head replacement on a given animal actor. "Our CG models were hand-modified or hand-built, based on photographs," says Visionart's executive vice president, Josh Rose. "Funnily enough, we were able to rework existing models we had created for commercial work and whatnot. We had several dogs and a raccoon model, which we were able to rework for a possum. We tried to reappropriate models for the goat and the sheep, but wound up having to build specific models for those characters. Then we mapped photographic textures back onto each character, which were either pulled from high-res scanned photographs of the particular animal or, in many cases, actually projection-mapped off the original plate, which was the only way we could make the effects appear seamless."
The key to creating successful talking animals, Rose insists, was keeping a real animal's eyes, even when much of its face was CG: "The eyes are so much of what brings them to life," he says. "On many of the close-up shots of Rodney the guinea pig, we ended up doing muzzle replacements; almost two-thirds of his head was 3-D, but we kept his real eyes. For the cat, the real eyes were also kept — the computer graphics feathered off about half an inch all the way around the mouth. The only animals we did CG eyes for were the sheep, because they were chewing so much and their entire bodies moved. We had to replace large enough portions of them so you wouldn't see our CG animation sliding over their real bodies."
One of Visionart's most challenging creations was a Spanish-speaking orangutan. Rose recalls, "Jon Farhat tended to find out what Betty and the producers were most excited about at the time, and then just embellish it. They were loving the orangutan, so Jon came in and grilled us really hard on that. Betty really wanted to see the sensitivity and the specularity and the rolloff of the light on those big, wet lips. The orangutan's whole muzzle area and the meaty part of his mouth were all 3-D. Stirling Duguid, our 3-D animation supervisor, built in all of these really insane little details — the crevices and wrinkles in the skin — and then added really big, sweeping meaty lips and a big tongue. He got a lot of articulation in there. There was a desire to give the orangutan a stoned, Cheech and Chong-style appearance, so Ted Fay, our 2-D supervisor, created that heavy-lidded look by warping his expressive eyes using Avid's Elastic Reality. In the end, it came out looking great."
Canada's Core Digital Pictures created some entirely 3-D rats, which meant tackling that age-old digital problem: CG hair. "Every single hair was a model," Farhat states, adding there was no other way to accomplish the task. "I had to shoot the trained rats at 48 frames per second with tons of light, but even then they moved very quickly and would not stay in a constant space. In order to keep them still, we had to feed them so they were always chewing, which meant we had to stop them from chewing to make them talk. Once we got into patching and fixing, it became apparent that the only way we were going to get through the job was to make three-dimensional upper bodies and track them to the lower bodies of real rats. It was one hell of a job, but we did it, and you can't tell the difference."
Simple as it may seem, Jon Farhat's ultimate goal on Dr. Dolittle was quite ambitious. "It had to look as if these animals were impeccably trained, not effected," Farhat says. "We didn't want to turn the film into a cartoon."
