How King Kong Works

In 1933, King Kong climbed the Empire State Building with Ann Darrow clutched in his hand and fought off biplanes circling his perch. It's one of the most recognizable motion-picture scenes of all time. The original King Kong was a series of mechanically controlled models, the product of stop-motion animation and a $650,000 budget. Seventy years later, King Kong is a 3-D model, the product of motion capture, digital animation and a $250 million budget.

The story is a classic: The beast falls for the beauty, and it doesn't turn out well for the beast.

Peter Jackson's "King Kong" still takes place in 1933 -- in one of his post-production diaries, Jackson says he couldn't pass up recreating the classic "biplanes vs. King Kong" scene that can only be pre-World War II. A team of scientists and filmmakers lands on Skull Island to track down a mythical giant gorilla. Beautiful actress Ann Darrow is along for the ride, and King Kong falls hard for the blond bombshell. The team uses King Kong's infatuation to capture him and haul him back to New York City. But the 24-foot (7-meter) ape breaks loose and sets out to find Ann. (Click here to watch the trailer for the movie.)

The computer-generated (CG) King Kong comes to life primarily through two techniques: motion capture and digital animation. The process requires heavy computing power: Weta Digital's cluster is one of the largest supercomputers in the world. Weta uses a PC Linux cluster running Intel Xeon processors. At the completion of The Lord of the Rings trilogy, Weta had 1,144 2.8-GHz processors. The company added another 500 3.4-GHz processors to assist in the final stages of post-production for "King Kong." The processing power used to create the giant gorilla and integrate him into the live-action footage equals the combined efforts of about 15,000 well-equipped PCs.

The work of animating King Kong took much longer than the live-action filming. For scenes that would includes both live actors and the digital Kong, the actors had to approximate their digital co-star with Kong-like stand-ins and props.

Motion capture and animation were happening simultaneously -- animators were busy building King Kong's skeleton and fur while Andy Serkis, the same actor who portrayed Gollum, was beating his chest in the motion-capture studio.

Motion capture is essentially the process of translating the actual movements of a live person into the digital movements of a computer-generated character. In this case, it was the process of capturing the movements of actor Andy Serkis and transferring them to a CG gorilla.

When you watch a scene of Peter Jackson's "King Kong" in which the gorilla runs up an embankment in the jungle, what you're actually watching is heavily edited and animated footage of Andy Serkis running up an embankment in a motion-capture suit. For all intents and purposes, Andy Serkis plays King Kong. He acted out each scene in a motion-capture suit, and his performance was sent to the animation department to be applied to the King Kong 3-D model.

There are several methods of motion capture, including optical, electromagnetic and mechanical. A mechanical motion-capture setup involves a heavy suit that is basically one big mechanical sensor. It relays data to the motion-capture computer when a part of the suit detects movement. An electromagnetic system involves a suit of magnetic sensors that receive signals from a magnetic transmitter. Each time they pick up a signal, they send their location information to the computer. "King Kong" used an optical system, which is simpler than either of these setups, at least for the actor in the suit.

This optical system consists of a lightweight suit, a lot of reflective dots and about 70 motion-capture cameras that feed information into computers running 3-D motion-capture software. The motion-capture studio is a huge warehouse converted into a high-tech space with bare-bones sets and more cameras and computer equipment than you find in an electronics store. The sets are simple structures meant only to simulate the kind of terrain and props that the digital King Kong interacts with on the screen. They don't have to look real, because the cameras can't see them. The cameras in this studio could only see the reflective dots stuck to Andy Serkis.

Serkis learned to act like a gorilla by studying them in Africa. He studied how they move, how they sit, how they run, how they beat their chests and how they roar. For his performance as King Kong, he wears a simple, blue lycra body suit -- there are no wires or sensors, which makes the suit lightweight so it doesn't hinder his performance. In addition to the suit, he wears arm extensions to make his arms more ape-like.

Once Serkis is in the suit, motion-capture technicians place about 60 optical markers on Andy's body and another couple of dozen on his face. The markers are just tiny plastic balls coated in reflective tape -- they're a lot like the kind of reflectors you stick on a bicycle. Each reflector marks a key position, typically a joint or a particular muscle. Several markers together represent a muscle group or a body segment. Motion-capture is particularly useful in capturing facial mannerisms -- for example, when a person (or a gorilla) looks up or down, it's not just the eyeballs that move. The eyeballs also drag the lids. It's a tiny distortion in the position of the eyelid, but it's essential for realistic-looking eye movement. Serkis' upper and lower eyelids are covered in tiny markers.

To track the movement of all of these reflectors, motion-capture cameras surround the set. There are 52 cameras that track the body markers and another 20 cameras that track the facial markers. Each camera has a bright ring of light immediately around the lens. These rings constantly send out bright light, and the markers reflect that light back to the cameras. In this way, each camera tracks all of the markers it can see at any given time. With all of the cameras feeding their data to the computer, the computer can triangulate the marker positions to create a 3-D "marker cloud" that represents the movement of Serkis' body in physical space.

But we don't want human motion in the scene. We want gorilla motion. This is one of the most difficult aspects of this particular motion-capture process: translating human motion into gorilla motion. There are some serious discrepancies in physical proportions. The software handles a lot of the grunt work in this translation. First, it maps the 3-D marker cloud onto a CG human body. For each part of the CG human body that has a marker on it, the computer determines a corresponding point on a CG gorilla body. The software translates Serkis' motion into the gorilla's motion as he is performing the scene. The end result is a real-time gorilla performance on the computer screen. Just like with a live-action performance, they shoot the scene until they get it right.

Before each motion-capture scene goes to the animation department, technicians edit the footage. Editing Serkis' performance entails tightening up the motion and doing touch-ups to make it feel more gorilla-like. This may mean lowering the brow to achieve a more angry appearance, adding transitions between facial expressions where Serkis momentarily rested his face or extending the swing of an arm where the cameras didn't capture the full motion.

A completed motion capture scene is only one step in developing King Kong as he appears in the film. When Peter Jackson approves a final motion-capture scene, it goes to the animators.

King Kong as we see him in the movie is an animated 3-D computer model. (To learn about creating 3-D models, see How 3-D Graphics Work.) King Kong's fur was the hardest to animate -- it's very difficult to make hair or fur look and move in a realistic way. Weta Digital faced similar challenges with Gollum, when they had to move each strand of hair on his head individually to coincide with the movement of Gollum's body. With Kong, the model is covered in fur almost head to toe. Getting the fur to move with the skin was an animating feat.

Animation is more flexible than motion capture. Animators alter and supplement the motion-capture performance when applying it to the 3-D model. They basically use Serkis' performance as a guide, laying it onto King Kong in the process of keyframe animation.

Keyframe animation entails programming key poses for a digital model that will create a framework for the model's movement. An animator typically works on a single scene until it's done, setting the exact position of King Kong's arm in one particular frame, and then the position of his arm a second later, so they look exactly right. The 3-D graphics software helps to fill in the motion that happens in between, creating transitions between each frame. In addition to body movement, animators also keyframe the face, pinpointing spots that they'll manipulate to achieve certain facial expressions. They heavily manipulate the eyes and the brow to make King Kong happy, angry, sad or confused. Using both motion-capture footage and their own research on gorilla mannerisms, they identify each point necessary to achieve a certain look and keyframe the stages in that facial expression or body movement.

Animating King Kong was not a sedentary task. The animators acted out each scene segment to get a sense of what it should feel like. They swung from doorframes in the Weta Digital offices. They role-played gorillas and dinosaurs and fought each other on the floor by their desks. For each second of each scene where King Kong is present, animators spent hours making it look real. They applied the motion-capture performance to the 3-D model, interpreted it within the context of the scene, acted it out, keyframed it, altered it and hand-animated every eye movement, nostril flare and lip twitch. In the end, the animators have translated Serkis' motion into King Kong's motion to create the actual digital performance you see in the movie.

For more information on King Kong, motion capture and digital animation, check out the links on the next page.