Normal Map - Smoothing groups / beveling edges
polycounter lvl 4
Hi polycomnunity, I have couple of questions but first I write what I know for better understanding problems. When we creating normal maps (in this case tangent space) for creating nice transitions in corners for illusion of beveled edge, we can have one or more smoothing groups , we can have one UV or have multiple UV pieces and LP can be inside or outside HP, or LP corner can be connected to HP middle of beveled edge.
Some examples:
First case, multiple smoothing groups/ hard edges and known problem on edges, when have hard edges on UV connected to others.
Second case show multiple smoothing groups/ hard edges with separated UV shells where all hard edges are not connected to others.
third and four case show one smoothing group / smoothed edges on entire LP with one UV shell and when all edges are sepparated, there isn't difference. We have nice edges but on sides have undesired and ugly transition. We can here add som beveled edge or supported edge loop for less transition between two sides of this box.
Conclusion: As we can see, first case isnt usable, second is best, and third and four case have specular problem on sides of boxes.
Last what we can do is right setup LP and HP and ofc Cage. And this is my question, what you use and what is better from these examples.
In this case we have on A1 LP corner exactly on HP middle of beveld edge and LP is inner of HP (smaller size), on B1 we have LP on same size as HP and C1 is outer of HP.
In this case I have topology for LP (hole in mesh), but I dont know what is better.
Some examples:
First case, multiple smoothing groups/ hard edges and known problem on edges, when have hard edges on UV connected to others.
Second case show multiple smoothing groups/ hard edges with separated UV shells where all hard edges are not connected to others.
third and four case show one smoothing group / smoothed edges on entire LP with one UV shell and when all edges are sepparated, there isn't difference. We have nice edges but on sides have undesired and ugly transition. We can here add som beveled edge or supported edge loop for less transition between two sides of this box.
Conclusion: As we can see, first case isnt usable, second is best, and third and four case have specular problem on sides of boxes.
Last what we can do is right setup LP and HP and ofc Cage. And this is my question, what you use and what is better from these examples.
In this case we have on A1 LP corner exactly on HP middle of beveld edge and LP is inner of HP (smaller size), on B1 we have LP on same size as HP and C1 is outer of HP.
In this case I have topology for LP (hole in mesh), but I dont know what is better.
Replies
http://www.polycount.com/forum/showthread.php?t=107196
basically a hard edge requires a UV seam, but a UV seam does not require a hard edge. Hard edges at UV seams are 'free' so it makes sense to do so in 99% of cases. If you want to use one smoothing group for some reason you're going to have to have a synched bake (that is, you bake in the same tangent basis your engine uses).
Anyways, usually the hard edges method is preferred, however there are some companies, who use their own solution and tools, like ID, where you have to bake with one smoothing group.
If xNormal and using Mikktspace, make sure "calculate binormal in pixel shader" is checked in the plugin settings. (i say this because i noticed some waviness on the flat surface of the cube).
For baking I use xNormal, target engine is UE4 and tangent calculator is ofc mikk tspace and in plugin have checked compute binormals in shader and waviness on 3 and 4 case is because I have 1 smoothing group for entire mesh.
My question is, when you create normal map and using hard edges, I thing if angle is more then 45 how you setup LP against HP, what you prefer (A1,B1,C1) and why. And from images (A2,B2,C2) when creating silhouette for LP what you choose.
What is this Mikktspace you speak of
Until it's very tiny bevels which cause extra problems
For an engine would be no difference actually.
Once following such pattern you could no more care much if baking synced or not. Otherwise that problem just could make you crazy
Going to try this next model for sure. Been concentrating too much on edge crap