Motion Capture

Motion capture translates to “movement capture.” A person moves, and cameras or small measuring devices precisely record where their arms, legs, and head are located. These measurements don’t produce a video, but rather a long list of numbers. The numbers describe the body’s position at every moment. A computer then transfers them onto a drawn character, for example a dragon or a cartoon figure. The character then moves exactly like the person who was recorded.

Why animators don’t draw every movement by hand

Human movement is astonishingly hard to reproduce. When walking, the pelvis tilts, the shoulders rotate slightly against the hips, and the head bobs almost imperceptibly. If an animator animates this by hand, a few seconds can take many days. And even then, the result often looks slightly artificial. Our brain recognizes incorrect movement immediately, even if we can’t say what exactly is wrong.

Motion capture solves this problem by simply measuring real movement. An actress performs the scene, and the data automatically contains all the small irregularities. This saves time and money. A game studio can record hundreds of movement sequences in a single day of shooting.

An important distinction: motion capture only provides the movement, not the appearance. Graphic artists still decide what the character will look like. That’s why the same recorded dataset can control a robot in one case and a bear in another.

From markers on a suit to evaluation by neural networks

The classic method works with a tight suit covered in small reflective balls, known as markers. Special cameras are placed around the room, often twelve to forty of them. Each camera sees the balls from a different angle. By comparing these viewing angles, the software calculates the position of each ball in space, usually 120 times per second.

From the ball positions, the program derives a skeleton. This skeleton is a simple framework of joints and bones, similar to a jointed doll. This exact framework is then fitted into the animated character. This step is called retargeting, because the movement is transferred onto a differently built body. If the character has longer legs than the actor, the software must adjust the stride accordingly.

For a few years now, it has also been possible without a suit. A trained computer program analyzes an ordinary video image and estimates the joint positions from it. Such programs have previously learned from millions of images what a human body looks like. The result is less precise than the camera method, but sufficient for many purposes. A common misconception is that motion capture also captures faces. Facial expressions are usually recorded separately, often with a small camera mounted on a helmet.

From movies to sports medicine

The technique is best known from cinema. Characters like Gollum from “The Lord of the Rings” or the Na’vi from “Avatar” were created this way. Major video games also use it for fight scenes and cutscenes. In sports games, the movements often come from real professional athletes.

Outside of entertainment, motion capture is a measurement tool. Clinics use it to analyze patients' gait after surgery. Coaches examine athletes' throwing arms to prevent injuries. Factories check whether workstations place unnecessary strain on the back.

In the news today, the term often appears together with AI. On one hand, learning programs are replacing expensive camera studios. On the other hand, recorded movement data serves as training material: robots and digital characters learn from it how to move convincingly. At the same time, there is dispute over who actually owns an actor’s recorded movement.