Surface normals and Geometry normals explanation

Nadra_Nagy

Hello guys, could somebody please explain to me the difference between surface normals and geometry normals simplified?
computational geometry - What's the difference between geometric surface normal and shading surface normal? - Computer Graphics Stack Exchange
I don't know how this affect the baking

poopipe

In the context of that post your normal baker will use the shading surface normal of both meshes.  
This means that smoothing groups on both the high and low res mesh have an effect on the resulting texture. 

Eric Chadwick

^ yep that's the answer you're looking for.

Also here's an app to help you visualize the differences between geometric normal and surface normal.
https://github.khronos.org/glTF-Sample-Viewer-Release/
Go to Advanced Settings at bottom right, then play with the Debug Channels.

gnoop

It's a terminology mess still existing in different 2d compositing and content creation apps.     What they call "geometric surface normals"  Blender calls face normals for example . It's mostly irrelevant  for baking. Just show you if some faces are flipped .   Honestly I am not sure why they need it in 2d compositing . Never needed it.   But I might be wrong.

"Shading normals" or what 3d packages call vertex normals  is only what's important .  And even here 3d packages call them differently . in 3d max it's just Normals.  You can see them if you apply "edit normals"  modifier .   Once  edited Max call them "explicit normals" painted green   vs default automatic  (blue) .

Blender call vertex(shading) normals  "split normals" and paint them pink.  While they could perfectly  be not split . Also only ones that are important . An actual vertex normals that do shading.  When they not split it's smooth shading . When split it's hard (sharp) edge.

3d max smooth groups is just a way to control vertex normals . Split them where necessary .

Normal map is like a cherry on a cake  working on top of  vertex(shading) normals but rather on per pixel level.  Basically adding  small per pixel shading variations to what  vertex normals are incapable to do when it's not enough  vertexes.