Smoothing Groups LODs Chamfer Normals Maps
polycounter lvl 18
I couldnt figure which thread to put this in since its talking about a cross section of all..
Ok here is something. Chamfer when you can versus using smoothing groups is what I take away from much of the discussion. Since smoothing groups when even able to be exported, are detached edges, so you end up with more vertices anyways.
BUT Here is something that I don't know has been addressed.
LODs.
If you chamfer an edge or put extra loops near it, they will be the first to close when the LOD goes down. So now your normals will pop between each stage earlier. VERSUS with a smoothing groups. These edges will keep their normal appearance longer in the LOD process.
In other words, that more pronounced groups of faces you emphasized with a chamfer versus a smoothing group will loose its display distinction earlier when dropping LODS.
So from this POV, it would seem Smoothing Groups would rule.
But the offset would be having to have a more uv islands for each of the smoothing groups where before you could have fewer islands. For burning/displaying the normal map correctly.
Soooo,, what am I missing? Other than the pain of the uv islands, it would seem that chamfering your low poly might be overstatted when dealing with the LOD. Right? Wrong?
Ok here is something. Chamfer when you can versus using smoothing groups is what I take away from much of the discussion. Since smoothing groups when even able to be exported, are detached edges, so you end up with more vertices anyways.
BUT Here is something that I don't know has been addressed.
LODs.
If you chamfer an edge or put extra loops near it, they will be the first to close when the LOD goes down. So now your normals will pop between each stage earlier. VERSUS with a smoothing groups. These edges will keep their normal appearance longer in the LOD process.
In other words, that more pronounced groups of faces you emphasized with a chamfer versus a smoothing group will loose its display distinction earlier when dropping LODS.
So from this POV, it would seem Smoothing Groups would rule.
But the offset would be having to have a more uv islands for each of the smoothing groups where before you could have fewer islands. For burning/displaying the normal map correctly.
Soooo,, what am I missing? Other than the pain of the uv islands, it would seem that chamfering your low poly might be overstatted when dealing with the LOD. Right? Wrong?
Replies
They were mostly just concerned with bitmap sizes for the most part. Well that and tri count in some circumstances I guess. Maybe their engine handles things differently but I have no idea.
If quality normals on LODs are a huge priority in your project, using more smoothing groups, and planing very carefully where you place them(to ensure you can do the same with each LOD) would be the way to go. You're still going to get discreprencies in some areas of course, but i think you can make this work.
Another solution is to do a new bake for each LOD mesh, so your normals are more correct, but this can be a little time consuming, and the new bake wont line up perfectly with your diffuse/spec textures(may not be an issue tho, depending on what distance LOD's pop in at).
At the end of the day, its an LOD model and may not be worth your time to spend too much energy in getting it to look correct, you should do some tests, and try to find a reasonable distance to pop the LOD in where any smoothing errors wouldn't be noticed, i think that is the best solution.
God, it sounds like object space would be my savior.... If the damn TGEA engine would support em!