SEB I couldn't seem to explain it well in words, so I recorded and uploaded a video that recreates the situation as shown in the ToomBoom video you mentioned:
In Spine, each bone and image attachment can have its own positions. In this video, there are three mouth parts, each with a different position. (In Spine, the translate X/Y value of an image attachment always indicates the position of the center point of that image.) The image attachments in this video have different Translate X/Y
values because the center point of each image is different due to the edited mesh hull, although the original images are all the same size. The same situation will occur if the original images themselves are different sizes. Each image attachment for mouth is a child of the mouth
bone, and when the mouth
bone is animated, it animates based on the position of the mouth
bone. In other words, the mouth bone plays the role of a pivot in ToomBoom.
I hope this is clear enough, but I think it would be easier to understand if you try the trial version.