Abstract: In order to let user watch the image about what he is wearing, a method for virtual try-on is designed using photos from the general model and from the ordinary user. The method includes three steps, the skin color substitution, the head substitution and the body shape varying. At first a mapping between the average skin colors of photos from the general model and the user is constructed. The color of pixels in the skin area of model photo is changed by the mapping. Then rough face boundary is determined by ASM, and it is further optimized to an accurate face boundary using the gray gradient. The hair area of the head is determined by diffusing from the initial hair pixels according to the statistical model of hair color. Finally, based on the model's pose ascertained by Kinect, the control points in the model's photo are obtained automatically. Using these control points the model's body can be changed to be consistent with the user's body. Experiments illustrate that our method is practical.