Help with High Poly
Hello, I've been creating models for a while now but I've never really learnt any High Poly techniques for Hard Surface modelling. With not having any work to do over the Winter period and the festivities not starting yet I wanted to try and teach my self.
So the problem is I'm creating a barrel (simple to start) and I created a clone with supporting edges and I applied the turbo smooth modifier, now as you can see on the left model which is the base with the normal map it's trying to create the roundness that the high poly created but it can't because of the more angular edges. I also created another clone with support edges running downwards down the barrel, these then made the normal map get rid of the smoothing group edges and show the edges. Although this did help with the horizontal sections.
if anyone can help and maybe point me in the direction of a nice High Poly tutorial to get me going. Thanks.
So the problem is I'm creating a barrel (simple to start) and I created a clone with supporting edges and I applied the turbo smooth modifier, now as you can see on the left model which is the base with the normal map it's trying to create the roundness that the high poly created but it can't because of the more angular edges. I also created another clone with support edges running downwards down the barrel, these then made the normal map get rid of the smoothing group edges and show the edges. Although this did help with the horizontal sections.
if anyone can help and maybe point me in the direction of a nice High Poly tutorial to get me going. Thanks.
Replies
http://wiki.polycount.com/NormalMap#Solving_Wavy_Lines
About subdivision surface modeling...
http://wiki.polycount.com/SubdivisionSurfaceModeling
We had a big thread about this a while ago, if you read through it you should find the answers you are looking for.
http://www.polycount.com/forum/showthread.php?t=72713&highlight=normal
Quick test to show various methods.
left to right:
A. Most accurate to high, wonderful, but insane tricount, 32 sides + all details modeled
B. Very accurate to high, except not enough sides. tricount quite high but has worst wavyness of all, 16 sides + all details modeled
C. Boring silouhette, but least wavyness outside of A, very reasonable tricount, 32 sides + support loop on either end
D. Boring silouhette, very little wavyness, very low tricount, simple 32 sided cylinder
Also note, in the last shot, at a reasonable viewing distance there is little difference between any of the 4 methods.
If the asset needs to be very high quality, seen up close, A. with some healthy LODs is the best method, if you're only going to see it at a medium distance, D is probably the best method, however ALL of these seem acceptable, the wavyness on B to me is well within acceptable margins from most angles.
I should have included a simple 16 sided cylinder example too, oh well. Lower than 100 triangles isn't really worth discussing(and arguably lower than 500) as the drawcall for the mesh is going to be what hurts performance, not a prop of such little geometry.
Rar with:
Highpoly obj(un-subiveded)
Lowpoly obj
Maya 2008 bake scene
normal map TGA
http://dl.dropbox.com/u/499159/barrel.rar
Now, general issues with your model:
It looks like your highpoly's details are too sharp and fine, resulting in a poor bake, the bevels on your low are also too small, leaving not enough resolution to correctly represent them in your texture, and just giving a very wimpy feel to the asset. And you have some smoothing errors, or possibly normal map applied incorrectly(does the green channel need to be inverted?) as well?
While in itself not useful it allows you to reuse the same normal map on the low-poly that you use on the high-poly.
I am also wondering if it is actually any more optimized reducing the tricount 892>764 when the vertcount remains the same.
Uploaded with ImageShack.us