Question regarding UVs and smoothing groups

Dethling

Hi,

At the moment I'm working on a scifi prop which contains a lot of beveled edges.
As my last models were all hand painted I aim toward a current-gen model this time to improve my skill set (polycount wise, normal maps, PBR etc.).

For the hand painted models I always tried to keep the UVs as much together as possible, so I can easily paint over edges and don't have visible seams.

For this project I read a lot about normal baking and I read that you should seperate UVs if they have different smoothing groups.

I only want to cross check if this is also true for < 90° angles.

Example:




Which of the three would be the best if I want to bake a high poly version (e.g. with some bolts, insets etc.) to the low poly version?

SuperFranky

I think A would work quite alright in your case. Hard edges are made to avoid extreme gradients on the normal map in places with 90 degree edges. In your case you should be fine. In any case, why don't you just do a couple of test bakes and see what you get?

Dethling

Thanks a lot.
Regarding the test bake: The model has a lot of these bevel edges and as it's my first complex model to bake, I thought asking would be better then "try, error > frustration".
I also don't have the high poly version ready yet, as I'm still working on refining the low poly and doing the UV Layout. (My Workflow: Blockout > Low-Poly > UV > High Poly > Texturing)

SuperFranky

Just remember, if you use A layout, then you should not use smoothing groups where there are no uv splits, or you'll get nasty black seams in your normal map

Bek

Essential reading: www.polycount.com/forum/showthread.php?t=107196

You say you've read that "you should seperate UVs if they have different smoothing groups" but have you understood why this is? Because that understanding is fundamental to this problem. Your lowpoly smoothing and UV seams are tied together; they both affect each other. When I work I usually unwrap, keeping in my mind that seams should be splits, and my spllits are inherited from my UV seams on export (what edges are UV seams become hard/split edges). Most of the time where you want a UV seam should be a split anyway, it's not a hugely complex process, which brings me to my next point:

The (non-technical) problem with your example is that it is too simple. If that were the amount of geometry being used you'd expect the asset to be very small/of little consequence. The better way to approach this (assuming it's an important/large asset) would be to add more geometry so that you can have minimal UV seams and minimal harsh shading. If the sides need to be continuous for texture reasons the top edges would be seams. If not you'd do either A or B depending on other concerns (what packs better, will the texture here need minimal distortion, what is more efficient, what normal map res I will be using (ie can I get away with less splits/harsher shading)).

igi

I tend to separate the surfaces where edge angles are close to 90. In your example I wouldn't spend extra time to separate the surfaces since these angles on your model are not that close to the 90. Once closer to 90, the severity of gradients increases and normal map are less likely to handle your surface data well. Case A should work fine. There's always expectations, sometimes it requires trial and error.

edit: gradients are not essentially a bad thing, it helps to compensate the surface differences between low and hi poly model. Normals are not good at capturing continuous surfaces with kinks closer to 90degree.