A perfect morph would require having one mesh for figure A and a copy of that to shape figure B while not adding or subtracting any points/polygons. This requires a good mesh of figure A with the idea of figure B already in it, with a mesh density needed later on.
A simple way would be to dissolve one object and do the opposite to the other, which often looks cheap and less attractive.
To understand what is needed, let’s look at image morphs. First, the two images need to be evaluated, which feature becomes what after the morph.
The eyes, nose, and mouth are often the key features of the head. So Image a starts to deform these features, so they sit on image B later. Image B moves those features to become A after a while. A to B and B to A, while that deformation happens, A is first visible, then blended over to B. IF that blend is global or progressively localized is based on art direction and other things.
So, how would that work for a 3D object? There are several options to shape A closer to B and B closer to A. each has a PoseMorph, and with some Polygon “shaping” (e.g., Magnet Tool), the middle where they switch from A to B can be achieved.
A rougher approach is via FFD:
Typically, those effects have some eruptive movements, where transitions can be camouflaged in motion blur. Especially when a beast comes out, it comes in a staccato way, and it is in most Werewolf Movies that I have seen the case. It often comes in a kind of a rhythm, so it looks believable, as the next eruption is predictable. If predictable, you have your audience often convinced.
Besides that, one could also use a method that “melts” two objects and covers the transition quickly. The Volume object can do that.
Other options come to mind, like the camera deformer to move AtoB, BtoA, or even the Collision deformer, where both get the same deformation at one point and (kind of the middle object method) create so the illusion of morphing.
All the best