pale rings on skin when exporting from Blender

[Lookn4Mods]

I have generally gotten the hang of Blender etc, but for some reason I am getting these pale halos on the skin of any model I export. Even on a straight import then export the halos are present. Most noticeably they go right around both nipples, but are also present in other areas like the hips. Does anyone know what causes this, or how I can get rid of them? Any help would be wildly appreciated, as I have tried everything I can think of, and searched and searched for an answer. See attached pics for examples (black is my effort to make work-safe).

 

 

[Tony the Wookie]

are you sure it isn't a problem with you texture, rather than the mesh itself?

[Lookn4Mods]

The rings are there regardless of what textures I use unfortunately. Same happens even if I use the stock .dds, _n.dds and _sk.dds. Texture doesn't line up at all, but the rings are still right there!

[Tony the Wookie]

But there is no sort of indention or anything with the mesh in that area... that is odd...

 

I don't really know, but the next place I would look would be you unwrap.

[Backsteppo]

Open the .nif file with NifSkope, then Spells -> Optimize -> Stripify all TriShapes.

[Lookn4Mods]

I've figured out the cause of the problem is the binormals and tangents. In all the Type 3, Type 5, GrowIf, etc models (and in all the outfits that use them as a base) that data is missing, and in NifSkope the "TSpace Flag" is set to "0". HOWEVER, in at least one body I've found (the Malos body) that data is present and correct.

 

The problem is, when I export a skin from Blender, it seems to automatically add the binormal/tangent data, which is wonky in the areas which show rings. I've figured out if I set the TSpace Flag to 0, then the binormals/tangents go away and so does the problem. Model still looks good, though not as good as it would if it had the data present and correct.

 

Which leads me to the eventual question: Does anyone know how to edit/correct the binormal/tangent data I'm getting out of Blender? I've searched and searched, but I can't find or figure out how to correct the data.

[Ghogiel]

the vanilla oblivion horses exhibit the same bad lighting cause by tangent space. I have a picture here somewhere....

http://i25.photobucket.com/albums/c85/lego...seArtifacts.jpg

 

first thing to try is updating tangent space spell in nifskope.

if that doesn't do anything, the issue could be in the UV. The skin shader in obivion could throw a fit if there were any stacked/overlayed UVs at all. and updating tanget space apparently wouldn''t solve. Which I can believe- I have a tutorial describing the old nut of tangentspace and mirrored overlayed UVs causeing bad lighting on the mirrored seam.

 

there is a old old topic on niftools forum about this as well.

 

Mostly its a lot of out dated posts that have missleading info for the unwarry, but may give some insight :shrug:

 

I suspect that stripifiying skinned meshes isn't optimal. Bethesda always uses shapes and not strips on skinned meshes. and I've wondered for years why. performance seem like the most obvious, and hence most likely reason to me. I'm not sure how that will affect the binormal

[Lookn4Mods]

That's it Ghogiel!!! That FIXED the data!!! Thank you so much, you don't know how long I tried to find a cure. That has fixed BOTH the tangent and binormal data too. You rock dude, I would kiss your feet, maybe even suck your toes a little. But only until it got weird and we both got uncomfortable.

 

The models look significantly better in game now. Better than the originals, because none of the type3, type5, etc etc meshes I could find contain tangent and binormal data. Presumably they too could not work out how to correct the values. Who would have know it could be so easy! Thank you again, really. I'm off to re-export my skins...

 

Thanks.

[Ghogiel]

:D yay.

 

a probably uselss experience- I wondered the same thing about the type3s. I while back I did a little look into it, I was poking at Backsteppos nuns&fathers- the screen shots had obviously nice normal maps, but the nifs had Tspace flag at 0. I found that a little intriguing.

 

I made a copy of one of the meshes and enabled the Tspace flag and updated the tangent space. in nifskope no Tspace flag negated the normal maps. but it seemed to me that in the modelviewer or GECK the normal map did display. there were discrepencys between the normals/lighting themselves, but I believe that during my tests I updated the tspace, which the original author did not. so the lighting was improved slightly. but the point is, even with Tspace flag at 0, the normal maps still worked. which was quite surprising to me.

[Lookn4Mods]

I noticed the same thing. The normal map is used, but differently. It seems to me the game "wings it" instead. The normal map is used more generally but more subtly, while additional shading information is added based on the mesh to compensate.

 

As an example, her ribs (which are on the normal map) are visible more often without the tangent/binormal data, but never very strongly. With the data present though they are shaded more correctly, sometimes completely invisible, and sometimes strongly visible, dependent on the light angle. Same goes with her stomach muscles. Without the tangent/binormal data the muscle accents on the normal map tend to give her a pot-bellied look, but with the data present the highlights show much more appropriately, flattening the look of her stomach again.

 

Other shading is different to, as without the data the game seems to try and gleem more information from the mesh instead. Most notably the mesh appears quite puckered around the neck seam, as if it is seeing the edge of the skin as marked sharp (not smooth like the rest), which feeds back to the next vertices along. But with the tangent/binormals present this puckering completely goes away.

 

Thanks again, it really is nice to have this roadblock behind me.