Started to learn about normal maps 2-3 days ago, and have encountered some problems that i can't fixed, gladdly, thei appear to be (SUPER) begginer tipe of problems, so may be easy to solve.
1. In this render, there are 2 ''problems'', one is on the corners, where my mesh ''stick out'' more than the lowpoly, but i don't know how to deal with it, should i tight more the edges? or should i change the topology? (its just a standard cube in subdiv)
The second one, may not really be an error, but me because i don't know how the normal really should looks like.
2, Where the edges meet, at the bottom of the cube, there seems to be a problem which i will call ''cut'', the normals highlight seems to favor only the edge that are more perpendicular to the view, and in the normal, this area is the only place where the ''top'' of the normal don't have a soft gradation. So, whats the problem here? the way that the uvs are placed could change the behavior of the normal in that part? or maybe is just my illumination that sucks.
This other porblem, happens quite a lot. For this cube, i used a 2k image with 16 pix for padding, and i end up getting that seam, i can fix that, increasing the padding to like 80, but that is way too much right?, so how to solve it? .The edges and vertex of the seam are highlighted with red in the normal map.
I have on more question not related to those problem, and is about anti-aliasing/super sample. When do you use it? so far i have only made tests with none anti-aliasing, and they seems too looks quite alright, so i don't know when i should use it, maybe the same rules of the ''most importnat objects in the scene'' apply to that decision too?
Replies
https://www.marmoset.co/posts/toolbag-baking-tutorial/
Be sure to read the Basics and Best results sections, as it covers many common problems. Your first issue has to do with the bevels in the highpoly mesh and the direction of the projection rays and is more or less expected behavior. The second issue has to do with hard edges and uvs (see the section dedicated to this in Best Results).