Hey, I'm having issues with seams in my normalmap and I just spend a whole day trying to figure out what is happening. I have baked-rebaked with different cage meshes, different fbx export settings, xnormal vs substance, hard edges vs soft edges on the lowpoly, locked normals vs unlocked normals, fbx vs obj. But, 6hrs later I still have the same issue.
Here's what happens:
https://imgur.com/a/qX9fDThe UV shells of the lowpoly are rotated pretty randomly in the UVs but nothing is mirrored/flipped/stacked. There's plenty of room around the shells for padding. Some shell seams are not/barely visible at all when looking at the material, but when you switch to normal view you can clearly see that the normals don't match up at the edge of any shell basically. Or rather, it's like the normals match up,c but there's a still a shading seam running along the edge of the UV shell.
And as far as I know it's not possible to clone-stamp normals over normal seams in Substance Painter if the shells have different rotation, so there's no way of fixing it really. If I paint 128,128,255 over the seam I can make it disappear.
Anyone have an idea of what is going on? I seem to get the same result no matter what I do or how I bake this mesh.
Replies
Given that all your bakes turn up the same artefacts I'd be looking at UVs.
It's hard to see exactly what's going on but if there's a significant difference in texel density at a seam then youll have trouble and if seams are at weird angles that will exacerbate the problem.
Try shattering the UVs into more shells with less distortion and see if you can ensure a good match in texel density at each seam. If that improves things then you know you're on the right track.
Fwiw though, i wouldn't be baking the high frequency details at all, I'd bake a clean high res down with just the larger shapes and then apply the small stuff in painter as height information.
You have tried flipping your green channel right?