Normal maps/Symmetry seam question

qwedeath

Hello guys, got some problems with normal maps using symmetry
I've been digging this question for about 3 days and got nothing
Here is some screens of my model:


Some test cubes(standart normals without baking):
- 1st cube i've mirrored by hand and welded 
- 2nd cube standart 3Ds Max symmetry modifier 
- 3rd cube simple cube with additional edgess
Wireframe shot:

Modeling done in 3Ds Max, rendered in Substance Painter, i've tried baking in xNormal and Substance and still got the same result. 

qwedeath

After some tests i realized that this issue is caused by overlapping shells on the UV map, but still have no sollution for this

Obscura

This topic was covered many many times here in the technical talk. But basically you need to offset your overlapping parts to like uv 1-2. But don't expect to get a 100% seamless result... Especially with a this glossy material.

qwedeath

@Obscura thx for reply, already tried that, not much of difference 
so basically the best tip is to avoid symmetry on flat/glossy surfaces while using normal maps?

Obscura

You should be able to get decent result using the offset method.

Macebo

This could be a issue with you baking the normals with the other half of the symmetrical model (low poly) not present. Just follow this tutorial, there is no secret for doing it: http://wiki.polycount.com/wiki/Normal_Map_Modeling#Mirroring

Obscura

That depends... In a case of an organic shape, yeah. But in the case of the cube or something flat with smoothing groups, it shouldn't be a problem.

But both sides can be thereduring baking if you use offset on the uvs.

Macebo

It actually is a problem... I modeled a car recently and my bakes would have seams even if there is shared smothing groups. As pointed by the wiki (and i believe it is true) once you apply a cage and push it away the normals can bent a little, and the rays can have no mesh to hit along the hole, generating seams.

Obscura

Cage pushes along normal direction so technically what you are saying is not possible. Getting lighting seams is possible but its not because of that it hits nothing. You get one if the another side is not present, because you'll no longer have the same normal direction as what if would have with both sides along the seam. Another reason can be related to the tangents or smoothing groups. In case of a head, you'd definitely need to have both sides there during baking, with one smoothing group. Or in the case of your car...same...because its not flat.

Qwedeath's issue is high likely related to his smoothing groups. Because he gets the seams even without a normal map.

Macebo

Obscura, what the wiki states is wrong then. 

Without a whole mesh i would still notice seams, even after moving the mirrored UV to +1 space... I encourage @qwedeath to try the method i sugested and tried myself. It works. 

Obscura

Could you please copy paste that part from the wiki, cause I can't find it. To be honest I didn't read through the whole page now but in the beginning it does suggest to delete the half of the lp. It says this in the "Typical Mirroring Workflow" part. - But as we know this won't always work... If it does say something like "once you apply a cage and push it away the normals can bent a little, and the rays can have no mesh to hit along the hole, generating seams." then yes, this is wrong. Or maybe you misunderstood something.

Having both sides helps in certain cases but its not always necessary.

qwedeath

@Macebo thanks for the link, i've tried that and still got shading error
I think its just me, maybe i've missed some setting or some
Link to archive with cubes: https://drive.google.com/file/d/0BzH0kx67E34EMi1aSTh3SDc1WjQ/view?usp=sharing
Export settings:
 
Baking settings:

UV Layout:

Result:

Obscura

I took a look into the mesh and the lp had weird normals. When I tried to select one, it selected all of them on one smoothing group. Reseting the normals fixed this, but it didnt solve the seam, its there even without the normal map. Then I made a fresh, new cube and used symmetry, no normal map. I got a weird result. I got a lighting seam, but not along the symmetry line. Instead, on the center of the the face of one side.


Then I made a test with a baked cube,  only one corner baked, so its  3x symmetry. Again, I got seams, but on weird places.



Now I would also like to know whats going on here. My screenshots are from Toolbag 2.

- Changing the triangulation direction changes the direction of the line errors.

Mind blown

Obscura

I did an another test with a simple plane with an edge on its center, and Ive split the UVs there, as well. The seam does not appear. But after using symmetry, it appears. So this is definitely something to do with 3dsmax and its symmetry...

qwedeath

@Obscura haha this is hell
Thanks for this research, i've also tried splitting plain and it also worked for me. Ill try some other things, maybe some experienced symmetry master will reply to this thread :pleased:
 

Macebo

@Obscura sure, take a look: "Sometimes an artist will decide to delete half of a symmetrical model before baking. This is a mistake however because often the vertex normals along the hole will bend towards the hole a bit; there are no faces on the other side to average the normals with. This will create a strong lighting seam in the normal map. It's typically best to use the complete mirrored model to bake the normal map, not just the unique half."  It's on that link i posted earlier. 

@qwedeath i got some results, using the same mesh you sent. I used Xnormal for this. The ONLY diferent thing i did was to make a cage inside Xnormal -- i didn't use the one you sent.  



I increased the glossiness as you can see. No seams (at 4k; probably on lower resolutions seams can appear)



I believe the problem was your cage in some mysterious way.
:awesome:
The normal map:

qwedeath

@Macebo thanks man but i am still getting this seams even with YOURS map, i dont know whats wrong with me  
Do you got seams on this cube without normal map? What do you mean under creating cage inside xNormal? Ray distance calculator? 

Macebo

I believe there is nothing wrong with you lol. Sometimes this kind of annoying stuff happens... What program are you using to visualize the map? Substance painter, Toolbag? Different softwares calculate normal maps differently.

I opened the mesh you sent in 3ds max, and believe me, there were no seams...
Xnormal has a tool where you load both the HP and the LP and you push the cage to cover the HP. Then you save the cage in .sbm format (internal to Xnormal) and use it. 

qwedeath

@Macebo strangely i got this error in all 3 apps: 3Ds Max, Toolbag and Substance Painter 

Macebo

In this case we should call the Ghostbusters because you got some poltergeist stuff going on there :awesome:

Let's summon the almighty @EarthQuake to rescue us. He knows a lot about this subject. I'm actually kind of new in all this; sincerely i have no idea what is going on not with you (you seems fine) but with your computer. 

qwedeath

@Macebo haha yeah this is what i am thinking about  
Maybe i am just sleepy and missing something, hope to fix it tomorrow 

Obscura

Macebo -  It says exactly the same thing as what I said. "there are no faces on the other side to average the normals with"  


Unfortunately manually adjusting the normal along the seam won't work because the cage pushes along average direction. But in a cave of a flat surface, the normal would point straight downwards. But yeah, half deleted mesh will rarely work. This should be changed on wiki I guess.
Also the seam appears even without a normalmap so testing different bakes is poitless because the error is related to the mesh. 

Macebo

@Obscura It makes perfect sense what you said now. I made a test with a completely flat cube. on the left side there was a "whole" mesh, with symetry applied. On the right, only half of the model i sent to bake. 



No seams, no issues at all. I guess there is no imprecision or chance of the normals bending in the case of a flat surface (the wiki implies there is this chance at all times, not only when the model is not completely flat).

The strangest thing in all this is that, when i opened the mesh @qwedeath sent, there were no shading errors at all, the mesh was perfectly fine.  O.o

qwedeath

@Macebo this is super strange, after today tests i learned that this is caused by inverted UV faces, but its not fixing if you apply symmetry.
This shading thing confusing me so much. 

Klo Works

I don't plan to open a new topic since this one there.
I want to know an answer if the symmetry will always show seam line ? 
even without normal map still show, it happen when you directly do symmetry to the object or invert the UV map.
some topics :  
https://forum.allegorithmic.com/index.php?topic=7917.msg38569#msg38569
http://polycount.com/discussion/173271/normal-map-seam-along-mirrored-uvs

I search a lot of topics and there isn't any solution , so better way to avoid it right ? 
but I notice a lot of people use symmetry in characters and it seem  fine  

@Froyok
said :
"I just loaded your low-poly mesh into Substance Painter, without baking the normal map from the high-poly I have already a seam if I use the "Bronze Armor" smart material. The seam appear because of the "Finish Rough" applied on the bottom layer of this smart material. I don't understand yet why and I will need to discuss with an engineer to investigate further. Meanwhile you can disable this filter and you should be fine."

  people mention this will happen if the material too glossy and metallic but if it mix or less gloss will not perfectly noticeable.

so is flat surface  not recommend while organic is ok ?

I really wish someone explain this situation .