Orthognathic surgery is widely used to correct dentofacial deformities and restore both functional and aesthetic craniofacial balance. Predicting postoperative facial appearance is a fundamental component of surgical planning, patient communication, and treatment outcome assessment. However, achieving reliable facial predictions remains challenging due to the highly nonlinear relationship between skeletal and soft-tissue structures and the considerable anatomical variability among patients. Over the years, several approaches have been proposed for postoperative facial prediction, ranging from landmark-based methods and statistical shape models to biomechanical approaches, including finite element models (FEM), mass–spring models (MSM), and mass–tensor models (MTM). More recently, deep learning has emerged as a promising alternative for learning anatomical relationships directly from clinical data.