An interactive training system includes computer vision provided by at least one video camera for obtaining trainee image data, and pattern recognition and image understanding algorithms to recognize features present in the trainee image data to detect gestures of the trainee. Graphics coupled to a display device is provide for rendering images of at least one virtual individual. The display device is viewable by the trainee. A computer receives the trainee image data or gestures of the trainee, and optionally the voice of the trainee, and implements an interaction algorithm. An output of the interaction algorithm provides data to the graphics and moves the virtual character to provide dynamically alterable images of the virtual character, as well as well as an optional virtual voice. The virtual individual can be a medical patient, where the trainee practices diagnosis on the patient.