Theoretical SUB-D


The biggest problem with the crease setting is that you will lose it if you remove the modifier or collapse the stack. Also, I’m not sure just how reliable it is if you keep editing the mesh beneath the meshsmooth modifier…

On the other hand, if it works for you, use it.


There is a crease setting for ePoly’s as well. :wink:


Ah, I never use the subd settings in the epoly rollout. I guess they work fine then :slight_smile:


In general I think I prefer the mesh smooth modifier as well. FWIW, I discovered you can set crease amounts in your epoly, and then when you put a mesh smooth modifier on the model it will obey your crease settings. So no worries about losing your creases that way if you delete the modifier or collapse the stack, which is nice. Still wondering though, are creases really all that bad, or can they be our friend?


farcry… for instance have made enourmas progress, there was a article about the bump thing, a texture that made more detail in an object, was pretty weird, dont know how it worked exactly anymore, caus it deformed some smoothed modifier ontop of it i guess.

i wont take to long anymore till we can make highpoly for realtime gaming.

zbrush can make a major roll in the future, looking at the millions of polys allready working realtime in there viewports etc.


Speaking of Zbrush (and subdividing in it), I just discovered a really great reason to model using proper technique… chamfered edges instead of creases I mean. OBJ and DXF (and 3DS I noticed too) formats don’t support creases. One option would have been to apply mesh smooth by smoothing groups, but Zbrush doesn’t support smoothing groups. So, no easy way out if you want hard edges outside of 3DS Max. Guess I answered my own question in the end, lol. :wink:


We are allready there.
Look here :
The Geforce6 with VS3.0/PS3.0 makes rendering AND very complex shaders realtime.
It’s not about the Polys, it’s all normalmaps. If you do not want to transform something, you can go and project it’s whole geometry on a cube. You will receive an object with just 6 sides and 4 virtices, but you won’t be able to see any difference between this one and the original object with it’s millions of polys.



Tom pawlik, i mean real high poly… this looks high, but it is not… used normal maps bumps etc etc.

i mean probably in a few years. a lot will be really high poly, so you can make all the nice deformations.

But on the other hand i have my doubt about the extend the game industrie will go to with modelling.
will there be a continues increase of polys in everything, or will it just stop after some point.

a lot of details can be mapped, but there will come a point that it wont matter anymore. looking at games 5 years back… and now… hmm, than i think that we will be at that stage 7 years or so for sure. or maybe 5…


Actually the question is, will we still be using polygons.

When a model is so detailed that the polgons take up only a pixel on the screen, it’s no use to represent them with triangles eh, you can skip that and just use the points to represent the model. A lot of research is going on in this area. points are added or removed as needed, like lod on meshes. It’s a bit like voxels but for surfaces.


Yes, I’m agree! Things will DEFINATELY have much to do with LOD and other sorts of dynamic poly-reducing.

Maybe even pixels will dissappear someday too? :slight_smile:


That’s very very interesting.

Do you have any aditional information about this?


It sure would be nice if 3d-graphics cards started supporting voxels

2 cents.


look at zbrush, that allready is skipping a lot, you just deform in stead of model by poly.
you deform hundreds of polys with one pull or push of a part.

it wont take long till we just wont need the displacement maps to get that detail in another 3d ab… maybe it will be all like that.


Piledotnet: there are many (technical) papers to be found, search google for ‘point cloud rendering’ or ‘point based rendering’.

here’s a site showing some video’s for example:

and another one, this one shows that the technique is interesting for interactive modifications:

moreover, volumetric effects such as clouds are inherently possible with the technique. Just like the buzz now is on unified lighting (think Doom 3), soon enough the focus will be on a unified geometry representation, like points. Carmack once said that’s an important aspect of voxel orcloud representations.

It even runs on pocket pc’s:
Imagine what would be possible if GPU manufacturers concentrated more on this then rasterisation.

there’s even a symposium on it:

some of their papers:

this site shows some complex scenes:
they mention the similarity with true volumetric voxels. Maybe point based rendering will be the death of the poly, as an inbetween step to voxels. On the other hand, all those techniques can of course coexist.

So the general idea is:

“Point based representations have been proposed as an alternative to triangles for 3D surfaces, with a number of advantages. No ’mesh’, or connectivity information has to be stored explicitly. This allows a simple and compact representation, ideal for fast rendering and editing. When combined with advanced rendering techniques such as splatting [ZPvBG01, Pau03], point based surfaces outperform triangle meshes in terms of rendering quality and data storage flexibility.”


true, … it already has been used i nthe movie industrie many times. but when you want to have really highres and good detail, than you have a lot of points, and it always has to be manually adjusted… or like in the matrix/spiderman, remodelled… and the scanned model is just used as reference.

i think zbrush is in some way point based, isnt it… ? it doesnt really store all the poly, and tri data.


Indeed, i started thinking Zbrush actually uses some of those techniques when looking up those papers. here’s a paper that is more related to an editing and texturing point could model application:

it’s actually a paper about an opensource application made to research those topics:

Heh, i suppose we have to appologise to hijack this tread, this isn’t really about sub-d anymore is it. Oh well, it certainly is teoretical. Maybe this discussion can be contuinued in a thread ‘The dead of the poly?’

edit: Actually, subd, zbrush and point clouds are pretty related if you look at the CySlice tool from cyberware; used to create subd surfaces from dense polymeshes, created from scanned point cloud data. there’s also a sample model to download with normal maps:


yea, i think its the same… they also mention brushes in the document… didnt read it carfully , just looked it over and saw it somewhere… would make sence that zbrush is the same… wouldnt know how the hell they could get away with so much polys etc

but oke this paper is from 2002 presentation… so it can be from 1998 or even earlyer that they developed it… so what are they making now…thats the question…

pitty this forum is allmost dead… no one has sub d questions i guess


Hey guys, I was reading this thread and came across celticdog’s method of creating holes (page 28). I tried it, played with it and out came this image:

Pretty cool, yesno? I like the way how it keeps the reflections quite neat (except the bit just right of the middle).

Meshsmooth: 3x
Total polygons: 95K
All quads: yea baby!

See the large image (1600x1200) here.

Aaaand I made an utterly cool desktop image (1600x1200) for all sub-d freaks out there :wink:
Get it here.

  • Rens


ah nice, I shall have to try celticdog’s method too :slight_smile:


hey! that one really cool looking sphere! :thumbsup: