• Blogs
• Tradeshows
• Partners
• White Papers
• Events

Big Picture

Dec 21, 2005 12:28 PM, By Michael Goldman

King Kong's Effects

Although Weta Digital started working on King Kong long before it wrapped work on Lord of the Rings: Return of the King a couple of years ago, the company's techniques and technology required a significant leap from its LOTR approach, according to Joe Leterri, the film's visual effects supervisor. LOTR production techniques were replicated and advanced in several areas, such as the method used to animate Kong relying extensively on actor Andy Serkis' performance on Weta's motion-capture stage. In other areas, such as hair and water/fluid simulations, Weta departed greatly from its LOTR methods, and, in some cases, started from scratch.

"Lord of the Rings was a good foundation for how to do the work effectively and creatively, and for managing the production process—getting it on screen and properly reviewed, and all that," says Leterri. "But in a few places, we had to upgrade—different hardware and so forth. But Kong also had a whole other set of things that were not required in Lord of the Rings, especially with fur and water. For this film, we had to re-think how we did fur, so we had a huge R&D effort to model fur, groom it, add dynamics and mud, shade it, texture it—all of those things. It took us about a full year to get it where we were happy with it—about six months longer than we planned. We also had far more digital environments than Lord of the Rings."

Eric Saindon, digital effects supervisor for Weta, says the company's hair solution, dubbed Weta Maya Hair, basically permitted the company to give Kong's five million individual digital hairs realistic movement. He says, "[Kong's hair has] great hair dynamics, interacting with things like mud and blood and leaves."

The biggest visual effects challenge, of course, involved animating Kong himself. Leterri says the project represents a significant advance on the way the company used motion capture to animate the Gollum creature in LOTR. This time around, the company still used its existing Giant Studios mo-cap system primarily for body capture, but added additional high-resolution cameras from Motion Analysis to, among other things, accomplish extensive facial capture of Serkis to drive Kong's features directly, rather than using facial capture merely as a template for animators to follow.

"This time, we got full body and facial capture, and Andy's performance drove Kong throughout," says Leterri. "For the facial capture, it was an expression-based system. The idea is based on a technique called facial action coating. That describes how muscles drive certain actions of the face. We used a version of it extensively with Gollum in Lord of the Rings, but there, our animators used it for reference and hand-modeled the expressions. Here, we built the system to drive the face, based directly on Andy's expressions, with 132 markers attached to his face. The system interprets his expression and applies that directly to Kong's gorilla face. We called it the 'gorillafication process.' When a gorilla smiles, its face can do things a person cannot do. So we had to design the system to read 'happy' on Andy's face and apply that to the movements of a CG model of a gorilla. It let us read the expression frame to frame to insert it into the performance of the animated character. Motion Analysis gave us cameras with high enough resolution to track the dots on Andy's face, and we had excellent data values that way."

Matt Aitken, CG preproduction supervisor on the film, adds, "We blurred distinctions in various ways, including having our motion editors take mo-cap data and clean it up and extend it in their own right. They were almost animators themselves, in the sense that they extended the data during the cleanup process as needed to keep the value of what was already there, but also to help permit a better performance out of the raw data." Aitken adds that, while Weta never attempted to mo-cap real apes, the company did use reference footage of gorillas to build matching animation of wide shots of certain movements, like the animal climbing up a tree, for example.

"We took the video footage and matched our close animation by animating directly on top of the video footage," he says. "We encapsulated the motion in Maya and translated it to our digital Kong. That process got our animators into the headspace of how gorillas really move, and it was a key aspect of how we animated him. Generally, though, we used motion-capture data as much as possible, and then key-framed on top of that as much as possible. But there were instances with dramatic scenes, like when Kong fights two T-Rex dinosaurs, where the action lent itself to keyframe, because a chunk of the movement involved things we couldn't do on the mo-cap stage."

Weta also advanced considerably its approach to digital doubles on the project—extensively using doubles for lead actors, particularly Naomi Watts, when her character is carried or otherwise physically interacts with Kong. Weta relied on a structured light-scanning system called Light Stage, developed at USC's ICT Graphics Lab, to assist realistic lighting efforts for digital doubles.

"We really raised our game when it came to digital doubles," says Aitken. "We made Naomi a digital asset herself, and that double gets extensive screen time, going full screen without any problems. We also put full clothing on some of the digital doubles, and got extra movement out of the cloth as they bend arms—techniques to bring them to life a little bit more."

Weta primarily used Maya for animation, Pixar RenderMan and Apple Shake for rendering, Digital Domain Nuke and Autodesk Discreet Inferno for compositing, and Massive for digital crowd work.

In all these areas, and many others, the sheer volume of data passing through Weta's infrastructure was often difficult to fathom. In the last two years, the company doubled its storage infrastructure and upgraded its proprietary asset management database system to answer these challenges. Weta also dedicated 4,400 Linux-based processors exclusively to the task of rendering, and had 130TB of online storage available during production, along with another 500TB of offline storage.

But Weta also had to factor into the equation the fact that Jackson's preferred method of working involved routinely changing direction on visual effects shots well after principal photography was underway. This led Weta to perfect its process of "post-visualization" (post-viz) to accommodate evolving changes.

Leterri says, "[Post-viz essentially involves] putting the layout back to the previz version to tell the story you want to tell, which means that now, more than ever, postproduction is part of the preproduction process."

In other words, according to Aitken, Weta needed to routinely bring proposed changes back to the previz animation team in order to have them re-visualize those changes against existing background plates in a quick and efficient manner as production continued unabated. This permitted Jackson to then edit them into the evolving version of the movie.

"Peter likes to work dynamically, and is constantly revising the form of the movie as he proceeds through production and into postproduction," Aitken explains. "So there can be a requirement for new or changed visual effects, created during principal photography, as Peter explores different concepts or rewrites the script. So that makes for shots or scenes where he will shoot the actors, for example, against bluescreen, and then, later, he will do a rough edit of that bluescreen plate photography. He'll then hand that over to the previz team to create what would have been previz if it had come before principal photography, with rough blocking animation of the shot using plates that have already been shot, rather than before they have been shot, as you would do in a typical previz situation.

"The goal here is to give Peter the material he needs to edit the scene—to do a fine cut on the scene. He would not be able to do that without previz style animation to cut to in the first place. Then, that fine cut is used as a template for us to do the full visual effects work on the sequence. This is something particularly relevant to Peter and the way he works, but it would be a very useful tool on any major visual effects film. It's another example of lines blurring during the course of production."

---

For a detailed look at the development of King Kong's color grading pipeline in synergy with the film's visual effects pipeline at Weta Digital, see the upcoming January issue of Millimeter.

© 2008 Penton Media, Inc.