Alright, so I am throwing my hat in and adding another thread to the long list of normal mapping related issues. I am at my wits end on this one. I have checked the polycount wiki, searched the forums, searched the general internet and while I have found many useful things, nothing has solved my particular problem. Here are the screenshots taken from marmoset, and these seams also show up in max on the Xouliul shader.



Here you can see my low poly, the hi poly, the uvs, everything. I have re-baked this thing out of xnoramls 7 different ways, and I even picked up 3ds max and learned how to bake out maps from there, same seam. I originally followed the logical method that I have always been using, uv half, mirror it over, and moved the uvs of the mirrored side off the 0-1. Since then I have tried all kinds of different combinations to get it right. Also tried looking up info on inverted lighting bugs, and that seemed to be caused by either the hi poly or low having its faces reversed before bake, which mine are not. So yeah, I am just out of ideas, if it cant be fixed then I will have to just uv the whole thing without mirroring and that will make my textures cry. Thanks for any help anyone can provide.

try removing all mirrored mesh and then baking just half the model. After you've baked you can symmetry the mesh again to get back what you got.

thanks for the input, unfortunately I have already tried that and got the same results. Which is weird because then you would expect the whole mesh to bake differently and not have this seam. =/

Could be low-poly normals, maybe not...have you tried playing around with the UV winding order and detection settings?.
http://www.autodesk.com/us/maya/docs...ent_Space.html

Do not mirror the flat planar shapes that you're getting issues on, instead remove the mirror edge down the middle, detach them from the main uv island and assing hard edges/smoothing groups to these areas.

The problem is an unsynced workflow, if you were using a shader/baker that was synced it would render perfectly, but with a broken tangent space you'll get these little smoothing errors and seams.

Generally I would avoid mirroring on shapes like this anyway(the flat planes, not the whole mesh) as mirroring there would be fairly obvious.

You could also bevel some of these hard angles to soften the lowpoly mesh normals and get better results, that might help.

The winding order thing looks a lot like what I was experiencing but I cant figure out any simple solutions to that if it was even that. Earthquake brings up a lot of good points and I think I will go ahead and just changes the model and it should bake out better. I thought having a line down the middle would be cleaner and work, but guess I was wrong. Out of curiosity Earth, when you are talking about a broken tangent space workflow you are just referring to taking models through multiple different software that all have their own little ways of reading tangents? Thanks for all the help.

Hey Turret, i think you may find this thread useful

http://www.polycount.com/forum/showthread.php?t=51088

Yeah, I read through that thread and thought it was interesting however my seams are not all straight like that. Also later in the thread people were talking about how if you bake it right in the first place you shouldn't need to do this, so I set out to figure why my bake was messed up. I have moved the seam over now so its not in the middle and along a transitional edge. It looks like its still there some, but its much much harder to see so I will probably just say the hell with it and move on. =P Thanks for the help guys.