Create Composite of Two Videos

Step 1: Carefully plan and execute the two shots to minimize the difference between the movements in the scene so each take has the actors movement and the camera movement the same in both takes.

Step 2: Depending on the shot determine which video will be used as the master then track, time, and stabilize the shots using the most efficient technique for the specific shot so that both faces tend to stay in the same position in the frame. It is easier to adjust the timing if the shot is made at double the normal frame rate but the comp is at the normal frame rate.

Step 3: When the shots are matched, use the most efficient method you can to blend parts of the images together. The exact technique depends entirely on the shot. 

Step 4: Remove the motion stabilizing you applied in Step 2. The workflow used depends on the method used in Step 2 but it usually involves parenting and tying the Position of a parent layer to the Anchor Point of the master stabilized layer.

I can't be more specific than that without seeing the shot you ware working with. There are just too many variables. You may also have to add in Rotoscoping and color matching. It all depends on the shot.