Abstract: In this paper, we propose a stable and efficient algorithm for animating mass-spring systems. An integration scheme derived from implicit integration allows us to obtain interactive realistic animation of any mass-spring network. We alleviate the need to solve a linear system through the use of a predictor-corrector approach: We first compute a rapid approximation of the implicit integration, then we correct this estimate in a post-step process to preserve momentum. Combined with an inverse dynamics process to implement collisions and other constraints, this method provides a simple, stable and tunable model for defonnable objects suitable for virtual reality. An implementation in a VR environment demonstrates this approach.