This paper presents a new method for modeling of a personalized 3D face model that has a hierarchical structure of the skin, muscle, and skull for realistic expression synthesis. We start from an accurate facial mesh reconstructed from the individual face measurements. The deformable skin model has a multi-layer structure to approximate different types of soft tissue and takes into account the nonlinear stress-strain relationship and incompressibility of the real skin. The 3D face model incorporates the skull structure which extends the scope of facial motion and facilitates interactive facial muscle insertion. Under the muscular force, the deformation of the facial skin is evaluated by the numerical integration of the governing Lagrangian dynamics. The resulting animation system enables realistic facial expressions to be dynamically synthesized.