Abstract: As the facial expression changes,the facial texture changes accordingly. To simulate this dynamic process of expression changes,this paper proposes a method of facial expression synthesis based on active appearance model. First,the relationship between facial expression and model parameters is obtained with an offline learning strategy. Then,the given person's facial expressions can be synthesized based on the learning results. In order to solve the problem of blurry eyes and teeth,a synthetic eye image and a teeth template are used to replace these blurry parts of textures. Experimental results show that facial expressions with different intensity and categories can be synthesized;the synthetic eye image enhances the realism of the expressions and makes it easy to implement eye animation.