The process is simple but very tedious and time consuming. Basically, you want your model to be as still as possible and take a video of the "action". Then you import that video to photoshop using "Video Frames to Layers" function.
Now you want to align all the layers to eliminate model's movement and possibly camera shake. After that is done, you just go through the whole bunch of layers and pick out the layer that has the best quality and expression. This layer will be THE static image in the final product.
Finally, you just photoshop in the moving parts from the other layers onto the previous above picked layer. The more frame your do, the smoother the animation :) Again, very time consuming but rewarded result.