hey bro! to achieve that exact effect you need to:
-roto the subject on a duplicate of the footage, precompose it afterward
-3d track the footage, create the camera from it
-apply pixel sorter 3 with multiple constrained masks on the subject (one mask on a different body part)
-duplicate the pixel sorted layer for each body part, then keep only 1 constraint mask for each
-enable 3d for each layer copy, change the anchor point of each to its actual body part center
-rotate slightly each one to achieve true 3d depth
-it works on my side, i hope this guide helps yourself as well, have fun bro! 🔥