Spatial keyframing for performance-driven animation

This paper introduces spatial keyframing, a technique for performance-driven character animation. In traditional temporal keyframing, key poses are defined at specific points in time: i.e., we define a map from a set of key times to the configuration space of the character and then extend this map to the entire timeline by interpolation. By contrast, in spatial keyframing key poses are defined at specific key positions in a 3D space where the character lives; the mapping from the 3D space to the configuration space is again defined by interpolation. The user controls a character by adjusting the position of a control cursor in the 3D space; the pose of the character is given as a blend of nearby key poses. The user thus can make expressive motion in real time and the resulting motion can be recorded and interpreted as an animation sequence. Although similar ideas are present in previous systems, our system is unique in that the designer can quickly design a new set of keyframes from scratch, and make an animation without motion capture data or special input devices. Our technique is especially useful for imaginary characters other than human figures because we do not rely on motion-capture data. We also introduce several applications of the basic idea and give examples showing the expressiveness of the approach.