The paper presents a cloth simulation system that can take large time steps, overcoming the numerical instability issues common in previous systems. The system couples a new technique for enforcing constraints on individual cloth particles with an implicit integration method. The cloth is modeled as a triangular mesh, with internal forces derived using a simple continuum formulation that supports operations like local anisotropic stretch or compression. A unified treatment of damping forces is included. The implicit integration method generates a large, sparse linear system at each time step, which is solved using a modified conjugate gradient method that simultaneously enforces particle constraints. The constraints are maintained exactly, regardless of the number of conjugate gradient iterations, which is typically small. This results in a significantly faster simulation system compared to previous methods. The paper also introduces a method for dynamically adapting the size of time steps during the simulation. The system is demonstrated to be much faster, with running times similar to those of Breen et al. [3] and Eberhardt et al. [5] for draping a 2,600-node cloth, and capable of simulating complex wrinkling and folding behavior on garments on moving characters.The paper presents a cloth simulation system that can take large time steps, overcoming the numerical instability issues common in previous systems. The system couples a new technique for enforcing constraints on individual cloth particles with an implicit integration method. The cloth is modeled as a triangular mesh, with internal forces derived using a simple continuum formulation that supports operations like local anisotropic stretch or compression. A unified treatment of damping forces is included. The implicit integration method generates a large, sparse linear system at each time step, which is solved using a modified conjugate gradient method that simultaneously enforces particle constraints. The constraints are maintained exactly, regardless of the number of conjugate gradient iterations, which is typically small. This results in a significantly faster simulation system compared to previous methods. The paper also introduces a method for dynamically adapting the size of time steps during the simulation. The system is demonstrated to be much faster, with running times similar to those of Breen et al. [3] and Eberhardt et al. [5] for draping a 2,600-node cloth, and capable of simulating complex wrinkling and folding behavior on garments on moving characters.