usually, the head’s blend shapes of a character are done before the binding process. there’s a way that allows you do this after binding, it’s quite straightforward: select all faces to be used for the blend shapes of your character’s head, duplicate them and move the shape outside the head and color it with some distinct color (just to be prevent a false selection). name this shape buffer_head and blend this shape back to the original head. if needed, add also eyes, teeth etc. to this buffer_head and group the objects (be sure you don’t change the group order as it will break the blend shape functionality). turn off all the unnecessary attributes of the buffer_head because you don’t want it to render at all. increase the blend node to 1 so that all changes to buffer_head are always visible in the original.
you can now export the buffer_head as a reference and blend all your face expressions into the head in the new scene. there’s an additional advantage behind this process: when animating you don’t need to go to the character’s original face all the time and change its expression, but you can just put an additional camera in front of the buffer_head and call it face_cam. just activate this face_cam whenever you want to change the character’s face and it will be reflected into the original this saves you some minutes everyday when animating.