[Solved] Mirrored UV Baking Issue and Visible Seams

Nominous

1. It's my first time mirroring UVs in 3ds Max and I followed the wiki page instructions closely. The first two screenshots are the low poly with just the normal map baked in Substance Painter, which looks great besides the circled, flat area. That area and the quarter circle extrusion are the only parts of the low poly that are not mirrored. In both the low and high poly (Quad Chamfer + Turbosmooth), that area shares the same smoothing group as the planar polys surrounding it. In the third screenshot, the low poly has Display Hard Edges checked and there are no hard edges around the problem area, although it and the circular extrusion have separated UV islands.

Edit: Both the low and high poly are contiguous meshes and they contain no stray vertices. There's a small floater above the top cap of the high poly, but I don't think it matters.

Both the low and high poly were exported as FBX files with Smoothing Groups, Tangents and Binormals, and Preserve Edge Orientation checked. Only the low poly has Triangulate checked. The UVs have 16px padding and I used Reset Xform on both the low and high poly before exporting. I checked the triangulation in Substance using the Polygon Fill tool and it matches the hidden tris in Max. In Max, the triangulation stays the same when I apply Turn to Poly with 3 sides. Does Turn to Poly automatically triangulate correctly or do I need to turn tris manually?

If I don't mirror UVs, the bake has no issues (last screenshot). What am I doing wrong with the mirrored version?




2. When a reflective material is applied, I notice a seam where the UVs are mirrored if I rotate the view just right. This also happens without the normal map applied and without Tangents and Binormals exported from Max. Interestingly, that problematic area from my first question has visible seams as well.  This thread has some of the best discussion on this subject that I could find, but the problem was left unsolved. Is it possible to completely get rid of mirrored UV seams?

Nominous

Solved this on my own. I just realized the normal information in that problem area looks exactly the same as the top and bottom edges of the quarter circle extrusion that joins the rest of the model. That area's UVs are offset by 1, whereas the corresponding polys connected to the quarter circle are not. This means the normal, AO, etc. information from only the latter polys are being baked and subsequently mirrored, resulting in that effect.

If anyone runs into a similar problem, I fixed this by switching the positions of those UVs and following these steps for inverting UVs. Step 5 caused an error, though, so I skipped it, collapsed the stack, and applied a new UVW Unwrap modifier.

Concerning my second question, I should have read further down the polycount wiki page for workarounds like detail maps, offset mirroring, and element mirroring. In the end, it seems that the seam can't be completely removed, at least not easily.