Abstract: A method of morphable linear fitting for facial expression synthesis is proposed. This method first gains shape vectors and texture vectors from given sample facial images of the same kind of expression. Then, based on the facial image morphable model, the linear combinations of the shape vectors and texture vectors of the unexpressive images are retrieved respectively using optimal linear fitting. Those combination vectors are kept in synthesizing the expressive image required. This method can be realized easily and the resulting image produced by this method is quite natural, lively and convincing. More importantly, this method can synthesize a facial image with mouth opened and teeth showed from an unexpressive image with closed mouth. It outperforms other existing methods for facial expression synthesizing.