Abstract: We propose a method for calculating deformations of models by deforming a low-resolution mesh and adding details while ensuring that the details we add satisfy a set of constraints. Our method builds a low-resolution representation of a mesh by using edge collapses and performs an as-rigid-as-possible deformation on the simplified mesh. We then add back details by reversing edge-collapses so that the shape of the mesh is locally preserved. While adding details, we deform the mesh to match the predicted positions of constraints so that constraints on the full-resolution mesh are met. Our method operates on meshes with arbitrary triangulations, satisfies constraints over the full-resolution mesh and converges quickly.