Normal map seam problem + artefacts in xnormal

Modelling, Textures, Animating and other general engine asset topics.

Re: Normal map seam problem + artefacts in xnormal

Postby WhiteDevil on Tue Apr 22, 2014 7:18 pm

To me it looks like if the model is in the same smoothing group and you get errors, then those parts which are making errors have different vertex normals.
So my suggestion would be that do "Edit Normals" modifier, select everything and hit reset and then do a xform reset.
User avatar
WhiteDevil
Been Here A While
Been Here A While
 
Joined: Sat May 28, 2005 1:09 pm

Re: Normal map seam problem + artefacts in xnormal

Postby marnamai on Tue Apr 22, 2014 7:30 pm

normal looks okay?
Image

it should be different smoothing groups for the bake, right?

does it matter if I change it after the back to 1 smoothing group for the game import or will that fuck up the normal?

Edit: just tried what you suggested Whitedevil, still the same results
User avatar
marnamai
Veteran
Veteran
 
Joined: Thu Nov 10, 2005 12:17 pm
Location: Belgium

Re: Normal map seam problem + artefacts in xnormal

Postby marnamai on Tue Apr 22, 2014 8:33 pm

Seems like its a problem with how the tangent is read by each software

Image

Image
User avatar
marnamai
Veteran
Veteran
 
Joined: Thu Nov 10, 2005 12:17 pm
Location: Belgium

Re: Normal map seam problem + artefacts in xnormal

Postby Surfa on Tue Apr 22, 2014 9:09 pm

Setting the diffuse to be the normal map will show seams no matter what, this is expected because of how a normal map works.

The reason for the seams when the normal map is set used as a normal map is because the way the renderer/viewer calculates the tagents/binormals of the mesh is different to that of the baker. This means that the information stored inside the normal map (tangent/binormal and normal direction) is inserted onto the model in-correctly, causing the lighting to create seams. UV-seams on your mesh mean that two different vertices are present. When the render/viewer loads the information from the normal map it applies the data to tagents/binormal that were calculated differently and so there is discrepancies at the point these two vertices overlap creating a seam. If you were to split the smoothing at the uv-seam the normal map baking produces a different result and the discrepancies are reduced (not completely removed). The way to remove these seams completely is to have a baker that have the same tangent calculations as the render/viewer.

Edit: Because object space normal maps work differently the tangents/binormals aren't used and therefore the result will look the same in every renderer/viewer. The hand-plane program takes information from an object-space normal map and outputs a tangent-space normal map based on the selected tangent-space calculations.
Surfa
May Contain Skills
May Contain Skills
 
Joined: Sun Dec 30, 2007 3:04 pm

Re: Normal map seam problem + artefacts in xnormal

Postby marnamai on Tue Apr 22, 2014 9:14 pm

thanks surfa, that was very informative :)
User avatar
marnamai
Veteran
Veteran
 
Joined: Thu Nov 10, 2005 12:17 pm
Location: Belgium

Re: Normal map seam problem + artefacts in xnormal

Postby Surfa on Tue Apr 22, 2014 9:22 pm

Decided to check out hand-plane again having not really looked into it in great detail and they have several videos on their site, which do a better job of explain these concepts (visual examples). I recommend you check them out.
Surfa
May Contain Skills
May Contain Skills
 
Joined: Sun Dec 30, 2007 3:04 pm

Re: Normal map seam problem + artefacts in xnormal

Postby MrTwoVideoCards on Tue Apr 22, 2014 10:18 pm

It also looks like you have zero padding on your normal map. On the main problem it's probably better to unwrap the legs as part of the body versus being separate shells. Smoothing groups and shell splits should have nothing to do with the normal map seam, assuming you import your model with the identical smoothing groups from which you baked them it should show up fine in whatever engine you're using to render them.

I would suggest padding out the normal map some more and see what results you get. Setting the horse model (the entire mesh) as a single smoothing group will also eliminate those seams assuming the padding comes out nice. 6=8 pixels is a good amount of padding to start with.
User avatar
MrTwoVideoCards
Monothetic
 
Joined: Thu Aug 02, 2007 11:18 am
Location: IN YOUR SOUL

Re: Normal map seam problem + artefacts in xnormal

Postby marnamai on Tue Apr 22, 2014 10:21 pm

those seams were there with only a single smoothing group, that normal in the op has 2 pixels padding, tried it with 8, same results,

all comes down to tangent space
User avatar
marnamai
Veteran
Veteran
 
Joined: Thu Nov 10, 2005 12:17 pm
Location: Belgium

Re: Normal map seam problem + artefacts in xnormal

Postby Stormy on Tue Apr 22, 2014 10:38 pm

I'd say re-unwrap your model and put the seams in less conspicuous places. Maybe cut the leg in half longways and have the outer flank of the thigh connected to the main pelt island. Looks like it's the only option.
User avatar
Stormy
May Contain Skills
May Contain Skills
 
Joined: Sun Nov 28, 2010 6:03 am
Location: Cairns, QLD, AUS

Re: Normal map seam problem + artefacts in xnormal

Postby Surfa on Tue Apr 22, 2014 11:35 pm

MrTwoVideoCards wrote:Smoothing groups and shell splits should have nothing to do with the normal map seam, assuming you import your model with the identical smoothing groups from which you baked them it should show up fine in whatever engine you're using to render them.


This isn't technically correct as if you use a tangent space normal map then the smoothing groups and uv-splits have an effect on how the engine calculates the tangents/binormal for display. The difference can easily be seen in marnamai post containing marmoset screenshots. Some engine allow you to import normal/tangent data export from your 3d program but this again will only work if your baking application also uses the exported data.
Surfa
May Contain Skills
May Contain Skills
 
Joined: Sun Dec 30, 2007 3:04 pm

Re: Normal map seam problem + artefacts in xnormal

Postby marnamai on Wed Apr 23, 2014 12:26 am

redid the uv's, baked a proper normal map in xnormal (8 padding, cage, correct swizzle)

gotta love it when it all comes together :)
Image

tommorow I 'll tune the sculpt a little and redo the diffuse (since it doesn't match the new uv's)
User avatar
marnamai
Veteran
Veteran
 
Joined: Thu Nov 10, 2005 12:17 pm
Location: Belgium

Re: Normal map seam problem + artefacts in xnormal

Postby Smurftyours on Wed Apr 23, 2014 5:48 am

Thats a really good micro pony
Image
jangalomph wrote:Wise words from a wise man. ^
User avatar
Smurftyours
Senior Member
Senior Member
 
Joined: Tue Aug 04, 2009 6:11 pm
Location: California, US.

Re: Normal map seam problem + artefacts in xnormal

Postby marnamai on Wed Apr 23, 2014 10:27 am

I can definitely see the similarities but its a little larger than that cute micropony :p

http://de.academic.ru/pictures/dewiki/66/Belgians.jpg
User avatar
marnamai
Veteran
Veteran
 
Joined: Thu Nov 10, 2005 12:17 pm
Location: Belgium

Re: Normal map seam problem + artefacts in xnormal

Postby marnamai on Wed Apr 23, 2014 2:43 pm

hmm, guess there is still something wrong after all ... rendered in xnormal x+ y- z+ , which should be correct for the cryengine, flipping green makes it even worse

Image
User avatar
marnamai
Veteran
Veteran
 
Joined: Thu Nov 10, 2005 12:17 pm
Location: Belgium

Re: Normal map seam problem + artefacts in xnormal

Postby Stormy on Wed Apr 23, 2014 7:10 pm

Dude that is not noticeable. This is the point where you trade off perfection for production time. No one will ever see that in game unless the game is all about close ups of horses asses. This model is done, put it down.
User avatar
Stormy
May Contain Skills
May Contain Skills
 
Joined: Sun Nov 28, 2010 6:03 am
Location: Cairns, QLD, AUS
PreviousNext

Return to Art Creation

Who is online

Users browsing this forum: No registered users