Animation data, from motion capture or other sources, is becoming increasingly available and provides high
quality motion, but is difficult to customize for the needs of a particular application. This is especially true when stylistic changes are needed, for example, to reflect a character’s changing mood, differentiate one character from another or meet the precise desires of an animator. We introduce a system for editing animation data that is particularly well suited to making stylistic changes. Our approach transforms the joint angle representation of animation data into a set of pose parameters more suitable for editing. These motion drives include position data for the wrists, ankles and center of mass, as well as the rotation of the pelvis. We also extract correlations between drives and body movement, specifically between wrist position and the torso angles. The system solves for the pose at each frame based on the current values of these drives and correlations using an efficient set of inverse kinematics and balance algorithms. An animator can interactively edit the motion by performing linear operations on the motion drives or extracted correlations, or by layering additional correlations. We demonstrate the effectiveness of the approach with various examples of gesture and locomotion.