Normals Issues - Tried everything,

Jowens

So I've made the Master Sword, from Zelda: Ocarina of Time, and I am having issues with my normal map baking, just around the blade section and a part on the hilt.

I gave myself a limit of approx 2k tris. and I have about 2200

I've tried everything but the shading on this area is giving me quite the headache.



Im pretty sure this is what is causing my issues but I have tried beveling edges to get a smoother transition, and adjusting the geo there but with no success...

Low Poly with Normal



As you can see, even with the bake it's still a problem.

Here is another problem area.



Im not sure why I am having these shading issues.

UV's:



Normal Map:



Low Wires:



Highpoly:



Low With Normal:



As you can see, that issue is not in the Highpoly at all...

I've played around with smoothing groups, beveled edges, etc. To fix the problem but nothing I do works.

This is the cleanest bake I have had yet, and have still had these issues.

I know this has to be a smoothing group issue, but I have no idea how to get the best bake with it still being a decent polycount.

The issue goes away when all of those edges are hard (obviously) but the bakes do not come out at all what they should...

I am about to the point where I just want to add some more geo and keep the normals hardened for shading.

If anyone knows what I can do to fix this problem please help, whether it be a smoothing group issue or if I should go about the lowpoly topology differently, or change my UV's.

Sorry about the long post, and huge images...

tristamus

Very tired at the moment but I'll try to help.

Looks to me like you've got an extraordinary pole happening there (Where all the edges are flowing to the one vertex). I'd say your problem is a combo of your topology and smoothing groups.

Poles & Smoothing Groups

respawnrt

Fix that pole by connecting the vertices that form paralel edges.You could try ading a suport edge loop all around those faces, since you don't have a smoothing group on them alone due to the ones that are on the blades sharp side.

Pola

I'd say the pole is definitely the problem as well and connecting parallel would probably fix it. However there might be another solution, its just a guess, vertex normals might be possible to manipulate them to fix the shading issues.

Maybe try passerby's script?
http://www.polycount.com/forum/showthread.php?t=99087

Although you might be able to just harden the faces only for that bit of the blade for that uvisland?

Dylan Brady

the pole is not the problem, you honestly shouldn't even have that vertex there anyways since there's no curvature its representing.

If you notice in the normal map, what should be a flat surface looks more like a cylinder. this is because the the bake is trying to compensate for the fact that you've smoothed the normals and they're angled outwards at the corners of the blade.
beveling the edge doesn't change the fact that you have angled normals, it just makes the problem less dramatic.
If you add a support edge instead (similar to a bevel, but doesn't add curvature)
it will eliminate the problem completely.

If you cant add more polys, you can try UV mapping the flat part of the blade on a separate island from the edges/cutting part and baking with the flat part on a separate smoothing group.

EDIT: GIF

Jowens

Awesome, thanks for the help guys, I didn't think about the supporting edge, makes sense now. Bake came out a lot cleaner this time, and I no longer have weird shading on the object.

Much appreciated all, especially the GIF.

Pola

Is supporting edges necessary or would tweaking the vertex normals of the flat part of the blade to all face forward from the surface also solve the issue?

passerby

that could solve the issue to, just get the face normal, than write it to the vertex normals around that face.

it's more or less what the support loop does, since once you get a support loop in, it makes the normals around the flat faces point straight up.

Mekhollandt

Not so much supporting edges as retopologizing the original with new edges to fix the poling problem. But to answer your question, no, that wouldn't fix the problem. Two many edges terminating into one vertex will just show up poorly pretty much no matter what. If that side of the model is perfectly flat, and you harden just those edges that meet there, maybe you could get away with it, though I doubt it. It's just good practice to keep things clean and avoid as often as possible more than 4 edges connecting to a single vert.