This article describes a method of constructing parametric models out of captured motion and skeleton data. Casting the problem as scattered data interpolation, our work is based on a multi-step approximation for the interpolation function with motion data compressed by principal component analysis. This leads to smaller storage and faster computation than those of previous approaches based on classical methods of exact interpolation. As a result, motion models can be constructed out of a rich set of example data, but can be used for real-time applications. We demonstrate a motion model controllable by attributes including those invariant for each individual, such as age, gender, height and weight. A parametric skeleton model is also constructed and demonstrated