The pose of the model at an evaluation time is determined by interpolating pose attributes of the unconstrained pose degrees of freedom from previously defined poses. A neutral value of an attractor function at the evaluation time is based at least on the pose attributes of the unconstrained pose degrees of freedom. Because of this, the attractor function is time coherent. Similarly, pose constraints are determined by interpolating from previously defined pose constraints. Inverse kinematics or error minimization techniques modify the pose attributes of the unconstrained pose degrees of freedom to minimize the difference between the attractor function value for the pose attributes and the neutral value of the attractor function. The final pose of the model at the evaluation time is specified by the pose attributes corresponding with this minimization.