With King Kong, visual effects supervisor Joe Letteri and the artists of Weta Workshop have raised the bar of animated CG characters to an unprecedented level. Unlike The Lord of the Rings’ Gollum, who played only a supporting role in the popular trilogy, Kong is a lead actor and carries Peter Jackson’s three-hour epic remake just as much—and really more so—than its human stars Jack Black, Adrian Brody, and even Naomi Watts. It is certainly Jackson’s obsession with the story that was the incentive for this development, but it is the efforts of Letteri and Weta that made Jackson’s vision a reality.To create Kong, Letteri looked back at the original 1933 film, but relied mainly on his research of authentic gorillas. “We started with a couple of things,” he says. “There was a design based on an albino gorilla named Snowflake in the Barcelona Zoo, who unfortunately died right as we were starting the project. We never got a chance to go visit him, but Peter and [special effects makeup artist] Rick Baker did visit him years ago, and liked him as a model because he had a very expressive face.”Preproduction was begun on certain scenes before there was a script. “We started off with the Empire State Building, because Peter knew that was a scene that was going to be in the final film. And we were sure it was going to be 1933. There was a rough design macquette of Kong, and we made a quick previs model, using Snowflake as the template. We had this old model of New York with the Empire State Building—a very low-res model that was built when they started to do the film in 1997. So we got going on that sequence, with a couple of CG airplanes, working out the previs of it.”In the meantime, Letteri and Weta devised ways to make Kong the most realistic CG character yet to hit the screen. Relying heavily on animal reference including videos and photographs, their research was aided by actor Andy Serkis (who also played Gollun in Lord of the Rings). To prepare for the role, Serkis spent time at London Zoo and also went to Rwanda, where he spent a week with mountain gorillas. “He brought back some incredible footage,” Letteri notes.The initial design for Kong used this material and the Snowflake macquette. The intention was to make Kong into a stylized gorilla kind of creature, Letteri says. “But that didn’t get us the Kong we wanted because he was in a way too designed, so we started to strip all that off and just go back to real gorillas and try to actually match the skeleton, the muscles, to get everything physically correct. This took about six months or so, and once we had that working, then we went back and added all the design elements back in, the broken jaw, the crushed eye, all those touches to sort of age him and make him who he is as a character.”Probably the most impressive aspect of the new Kong is his face, which is totally mobile, with none of the stiff or sluggish areas that spoil the expressions of previous CG characters—even Gollum’s in certain instances.“We started off with a whole facial animation system that was based on what we had done with Gollum but a little bit more sophisticated, and in some ways streamlined from what we had done on Gollum,” explains Letteri. “One of the things we were able to do was figure out a way to motion capture Andy Serkis’ face, which we couldn’t do for Gollum, and what was unique about it is that rather than just a kind of retargeting thing, where you take the points on Andy’s face and move the points on Kong’s face, we came up with a system that would actually allow us to interpret and translate expressions.”To achieve this, the software read the entirety of dots on Serkis’ face and then calculated the movement of the underlying musculature. From that data, Letteri and the Weta artists could determine where a given expression was in the spectrum of expressions Serkis could make. Next, they would find Kong’s expression in the expression spectrum they had created for him, and match the two. It was in this way that Serkis could actually physically drive the performance.“We only used that for about 20 percent of the shots; we had 60 animators working on Kong, so there was still a lot of keyframe animation done, but what it did was it helped to jell the character, because we started to look at the choices Andy was making in performance. He had performed with Naomi on set, and then came back and did this in the mo-cap, so it gave us the lock on the character—it gave us that sort of actor’s mentality on what Kong would be doing, and that served as a focus and a jumping off point for all the animation that we did.”One of the main problems plaguing the original King Kong was Kong’s hair, which moved with every touch of the animator’s fingers, creating a bristling effect. And in recent CG films, such as the Disney version of Mighty Joe Young, hair still remained a major impediment to realism. But with Kong these problems seem to be solved. Letteri and the Weta artists basically modeled each individual strand of Kong’s hair, using a combination of procedural tools and computerized hand-grooming tools, in order to make the hair flow the way they wanted. “But you also have to deal with all of the physics of the hair, the dynamics,” Letteri says. “So we took it to the next level, in that instead of dealing just with the surface of the hair we actually dealt with how light is transmitted through the hair, because that’s one of the complex effects you see with real hair. And the thing that’s really tricky about it is that you can’t separate the two, because the hair defines the surface, and in some ways the surface defines the look of the hair. So it was a real back and forth process, to dial the hair a little bit, then change the groom, because if the groom is smooth the hair looks sleek and shiny. You can take that same hair though and rough it up and it looks really coarse. So finding the right look was a combination of finding out how coarse the hair was, and how shiny the individual hairs were. That was a good six months process to figure that out, too.”While working on Kong himself, Letteri and his crew were also creating the spectacular environments for Skull Island and for the New York sequences seen in the beginning and at the climax. “We did very extensive previs. Christian Rivers, who was Peter’s storyboard artist and art director going back a long while, wound up doing previs on this because Peter decided not to storyboard anything. We had these big action sequences, the Empire State Building, the Kong/T-Rex fight, the Bronto chase, and Peter knew we needed to get started on them early because they were so big.”The previs was designed and built to the proper scale, so instead of simply being used a guide, it was treated as the first pass of animation, a process that has its roots in the work Weta did on Return of the King. As part of the follow-through of this process, Christian Rivers became the animation director for the film.To create vintage 1933 New York, Weta obtained a data set containing New York city as it is now, from which they removed all the buildings built since 1933. Then, having studied the architectural styles prevalent in 1933 New York, they built a huge library of details—window frames, light and telephone poles, park benches, cornices—anything and everything to make the virtual city visually complete. “We built the whole city,” Letteri says, “It’s all there.”They then wrote software that could create entire neighborhoods based on given data requirements. “We’d run the software, and go into the city neighborhood by neighborhood and describe what kind of neighborhood it is, brownstones or whatnot, and get it to build it for us. It was very accurate, because we just kept going back to the historical ph
otos and matching them until we got it looking pretty close.”But the work on New York was not all CG. There were also large-scale set pieces, including a full-size set for Times Square that also doubled as Herald Square. “We only built to the first level. We always knew we could extend everything,” Letteri explains. The set was created to match the CG work and vice versa, so it could literally be inserted into the CG environment with comparative ease. “We could just drop that set in the middle of our virtual set; just overlay it on Times Square, and then whenever we had a camera track we would just shoot off and render what was up above the buildings that were on the set or what was down the street. It all fit together pretty tightly.“In New York, all the aerial views are completely digital,” Letteri says. “Skull Island was a combination of foreground sets, miniatures—mid-ground and sometimes it was foreground miniatures, sometimes it was digital extensions, sometimes it was completely digital; it was a mix of techniques depending on what the camera moves were and what the script demanded.”Skull Island presented more complex digital problems, with scenes such as the Kong/T-Rex fight happening on numerous virtual locations. “Kong and the T-Rexes go from one cliff to the next and that sort of thing. What we did is that we took the geo that was built for previs, and we found the area where any critical contact would occur—where the T-Rex jumps from one level to the next, or smashes into a tree, and we called that area the red carpet. We took that and we did one-meter contours for the art department and smaller sizes for the miniatures unit. We just spit out slices on a big grid and they used those as templates. So that way the art department had everything full size for a bit where maybe Naomi had to run down and across a patch of ground; that would drop back into the previs, that terrain; and the same thing with miniatures; they would build terrain and then we’d copy all the camera moves and shoot on the miniatures, and then the terrain would fit back underneath it, and the camera moves would go where the previs were. The two would lock together that way.”
Written by Henry Turner
