Summary This paper provides a solution to the problem of animating a character using motion capture data from an actor with dissimilar size and proportions to the target character. An additional goal of the paper is to solve this problem while meeting the online and real-time demands of computer puppetry. The authors' method utilizes an approach that attempts to dynamically identify the most important aspects of the motion (root of the character, joint angles, end-effector positions) to map to the target character. Key Ideas 1. Kalman filters are used to filter out noise in the motion capture data. This is important since capture systems that provide real-time output are very prone to signal noise which must be dealt with in a fast and efficient manner in order to preserve the real-time requirement. 2. Determining important aspects of a captured motion is based on a series of guidelines: Root position and end-effectors that don't interact with the environment are typically unimportant. It is important to note end-effectors that interact with other objects in the environment, including its position leading up and to the interaction and moving away from the interaction. 3. The Inverse Kinematics solver estimates the root positions and computes the body and limb posture based on the important aspects of the motion. Contributions a) Importance classification system that identifies which motions to map to the target character b) Inverse Kinematics solver that incorporates the end-effector importance information and meets the real-time constraint.