Modelling human-like characters, and animating them is difficult, in particular if you want extremely accurate (and or complex) motion. Motion capture ('mo-cap' amongst friends) is a powerful technique for recording motion data for animating humans (and other animals).

A motion capture system consists of a collection of sensors that simulataneously feed position and orientation data to a computer. By carefully mounting sensors to a human body, the human's movements can be recorded in real time. The recorded data can be played back by channelling it to a virtual character. If the position of the sensors on the human and their equivalent positions on the virtual character have been configured correctly then the results should be very good.

The two most common ways of performing motion tracking are magnetic and optical tracking. While wireless magnetic tracking systems have now come on to the market, magnetic tracking systems usually require cables to be attached to the body of the human, which restrict movement. Optical tracking systems involve reflectors and a set of cameras, but while this gives the human full freedom of movement, reflectors can become obscured so that data is not captured. In contrast, magnetic trackers produce data continuosly.

Noise filtering is an important issue when doing motion tracking. Magnetic trackers, in particular, by the presence of metal objects in or near the magnetic field and are not 100% accurate so that some form of smoothing and/or filtering isrequired to improve the quality of the recorded data and remove erratic motions.

The placement of the sensors themselves on the human is also important. Too few sensors may not give good enough results, while too many sensors can quickly produce too much data. Also the virtual human must be well designed using a hierarchival or inverse kinematic model so that it can be realistically animated using the sensor locations (on the body) and the recorded animation data. Creating correctly-proportioned human characters is not very easy and it is important that the data from the real world is scaled and offset correctly to fit to the proportions of the virtual human.

Optical motion tracking can also be used quite effectively for capturing facial expressions, however quite a few sensors (and therefore a lot of data) are required to capture even the simplest facial expressions well.