My impression is that the 3D modeling in the Blender version has many things that are too small and/or detailed for the output resolution, and therefore look qu

My impression is that the 3D modeling in the Blender version has many things that are too small and/or detailed for the output resolution, and therefore look quite bad, whereas the hand drawn version prioritizes clean pixels and readable shapes at the expense of "realism": - the rims of the stacked drinking cups - the rims of the stacked plates - the filling of the pie - the shadows under the plates, the pie and the mini-billboard Some objects in the Blender version are also a bit off regardless of the output resolution, for example the huge pipe and the wide forks. I think more iterations of improvement on the Blender version would have allowed results close to the hand drawn version or even better. For producing many 1-bit illustrations of this kind for a whole game, mixed techniques are probably the best path: retouching imperfect NPR renderings, assembling drawn and rendered elements in an image editor, using pixel art as a reference for Blender and hand-drawn sketches as a reference for both drawing and 3D models, applying filters to photographs, and so on.