back to page 3

One particular action that was spontaneously performed by Williams occurred as the production was shooting the Flubber discovery sequence. Karpman explained, "Robin thought it would be cool if he pantomimed stretching the Flubber and then sticking

"How in the world are we going to do that shot?" Tom Bertino, after watching Robin Williams improvise a movement with the non-existent Flubber

his face into it. Well, Tom, Keith [Johnson, matchmove supervisor] and I were on the set watching this... and our jaws hit the ground. We both had this expression on our face, saying 'how in the world are we going to do that shot?' It wasn't planned or rehearsed, it was just something Robin did, and it was such a great idea that Les wanted it in the movie, and we knew we had a great challenge ahead of us."

Bertino said that ILM built a dummy version of Williams in the computer, using primitive shapes to represent him in the 3D world. "We took measurements from the set," Bertino said, "like the distance from Robin to the camera, but because of the frenetic nature of the action in the scene, they're only partially accurate. It's not like we were matchmoving an inanimate object, such as a camera movement to a table, or something. A lot of it comes to our matchmovers eyeballing, trying to match frame by frame Robin's actions in a 3D world."

The Flubber model for that sequence was particularly difficult to create, due to the complicated nature of the face impression. "The challenge was to get that 3D impression of Robin's face, over Robin's face, and have his face refracted through it in such a way that it would be recognizable," Bertino noted. "That was the trickiest part--making sure that his face wouldn't be distorted so that it would be unrecognizable or disgusting. The rotoscopers had to deal with putting the Flubber in his hands, of course. From the time we got that background plate scanned, the shot was in production at ILM for three months."

Bertino said that a 3D cyberscan of a bust of Robin Williams was made, "but we didn't end up using it. Tony Hudson, one of our very talented animators just built a caricature to work, just as a part of a test concept shot, and it was so good that the director said, fine just go with that."

Visual effects supervisor Peter Crosman and ILM co-supervisor Sandra Ford Karpman go over a shot, while sitting in the middle of the oversized set, built for the Flubber dance sequence.

In addition to the shots where Flubber interacts with Williams, ILM was tapped to produce the dance sequence, where Flubber splits off into several Flubbers, and goes into a complicated dance number. Bertino said, "The salient characteristic of the Mambo sequence was that the backgrounds were all shot here at ILM. We built a huge oversized set that was three times normal size, including the whole window box, the plants, even the National Geographic magazines. To get the kind of sweeping camera moves we wanted, we couldn't do it on the real set--with a larger set, we could move our camera much more freely."

Karpman added, "We had a lot more control over those scenes, because the Flubber was the star of the shots. We had a great deal of input into the lighting of the shots. Ray Stella, who shot the second unit, was an absolute joy to work with on that sequence." She continued: "I wanted the camera to be choreographed in such a way where no one could say, 'oooh, that was a neat CG or motion control camera move.' I wanted the camera to move naturally and not robotic." Karpman said that camera moves were first designed in a 3D system, 'photographing' rough 3D animation of Flubbers going through their basic motions. "Instead of just flying the camera in and out of positions, we followed the physical attributes of real camera motion. We tracked the camera linear motion, then panned and tilted... we wouldn't arc the camera in these unbelievable motions that simply cannot be achieved with real production cameras," Karpman said. "The computer camera movements were then ported to the motion control camera, which shot our background plates on the oversized stage."

Rendering the Flubbers for the dance sequence was a tricky situation. Bertino explained, "We had the difficult problem of dealing with a transparent character, which isn't usually tackled in the world of CG character animation. You can't see through a dinosaur or a bug alien--they're solid, tangible objects. One of challenges throughout the entire movie was getting the reflections and refractions of flubber accurate. In the mambo dance, the problem was multiplied by a gazillion, since we weren't dealing with just one Flubber... there were Flubbers all over the place, moving next to each other, going behind each other... and they all had to look believable. It was quite a task just to get those shots to render."

Karpman said that ILM used a new ray tracing renderer to produce the Flubber elements: "It did a terrific job creating the refraction distortion of the Flubber. It really came through for sequences where there two Flubbers in a shot."

go to page 5