Theoretical SUB-D


urg: thanks buddy! we can’t let this thread slip too far down the page, eh? :wink:

ian: edge extrude is great for control, but for certain things (specially non organic) it’s nice to be able to take care of things all in broad strokes, to make sure the accuracy/consistency is there without days of tweaking. like anything, i’m sure deciding the method based on the situation is probably best.

for example [i haven’t modeled a car, but] it seems like edge extrusions would work great for the body, and lots of mechanical actions would be best for the wheels etc.


Yeah your right, it is situation dependant… I just find it a little easier sometimes when I don’t have to think about the whole volume constantly.


yeah i know what you mean. i’m starting to think that i should start using edge extrude when modeling faces and let the edgeloops evolve, as true box modeling seems to be a more advanced method that requires you to know exactly where you’re going with it. bit intimidating at first.


jum’bok, check out the modelling forum here at cgtalk. There’s a link to a really detailed modelling tutorial that deals with the poly by poly method.


julienj’s dobby tutorial right? yeah that was just amazing! i was really stunned how good the likeness was.

i’ve got so many tutorials on my harddrives and favorites list, now i just need to actually do something with them hehheh :wip:

hrmm i was hoping this thread would perk up a little bit. i wonder if 3dz is still planning on making this into a pdf file some day. i hope he downloaded all the images, because the earlier pages of the thread are starting to get link rot.

anyways, back to class for me.


this is possibly off the current Sub-D topic, but i found a wierd trick.

Take a model that has some fine detail and mesh smooth it

take a duplicate, convert to sub-D and the back to poly with Adaptive option box check.

Apply the same Hard/Soft normal option to them both. And the Sub-D conversion has cleaner shading in the detailed areas. I have no idea why the mesh is identical as far as i can tell and both have the same hard/soft option.

This may be just with me, try it and let me know =P

*edit - yeah BPT is god. =D


Hmm, I guess that applies to maya yes?


yes maya. sorry didnt say before! =D This is a sdsmax thread lol! found it in a search. O well already posted didnt I =P


hhm… cant it be that when you ad a smooth modifier on your work mesh , it hasnt go the precise co ordinates anymore, but when you have smooth and ad a edit thing over it again(dsont know maya) it gets the precise coordinates back, so that the primary light source look better (shade)


that might be it, it is esecially noticeble on a tight crease, I know one thing, i have retired the mesh smooth, and started using Sub-D and convert back with 1x division. The lighting is much crisper.


wow! this thread is so ridiculously helpful! I’m gonna start posting my current character stuff here soon.


Just read about this a page backa and it has been an issue thats had me wondering for some time now… why does game geometry precise specifically only tris?


Basically because everything you do will be stripped down to tris when it comes to the engine and the gfxcard. So, by working with tris you can optimize your model/scenery better. Although these days the gfxcards can handle so many polys that it’s becoming a non-issue. Only (well not ONLY, but almost only) rts’ games still need fairly lowpoly characters these days. FPS, fighting games, sports games… Even mmorpgs have fairly large polycounts.

Anyway, it was more of an issue back in the days, it’s becoming less and less of an issue now :slight_smile:


Originally posted by urgaffel
. Only (well not ONLY, but almost only) rts’ games still need fairly lowpoly characters these days

these days are over, well what do you call “fairly lowpoly” 1500-3000 Polygons for a charakter (as in our game) is pretty much high i guess or do you mean 7000+ polies? It all depends on how good your LOD is :wink:


I was thinking of the LOTR rts games and such, they average a couple of hundred polys for each unit (I think… Definitely less than 2000 ;))


ok you are right, but games like the lotr game or Rome:Total War are designed for massive fighting scenes, i guess if you want a scene with a few thousand units in a shooter you’ll need pretty lowpoly units too :slight_smile:


Its also in truth, how the game engine reads the polys. Whether it can draw true quads. Generally most can only draw tris. Also there are engines out there now that can read patch data, i.e patch models. The great thing with this is that its got inbuilt LOD.

Also you have things like critech’s, polybump mapping tools which are amazing. Which in affect bake detail on, through a texture.

The key thing is that its gunna go into the engine as tris anyhow so you might as well model with them.



I’ve released a time lapse video of a 15 minute study using the TubeCap primative.

You can check it out here:



Originally posted by CarlCampbell
Just read about this a page backa and it has been an issue thats had me wondering for some time now… why does game geometry precise specifically only tris?

well, we’re talking about realtime engines - and thats allready the answer to your question, realtime. what does realtime mean? that the engine that is running is able to render the geometry in realtime - so, what’s the most simpel type of geometry? verticles and triangles - they’re simple, and thats the reason why they’re rendered fast - thats why game engine coders use them… best possible quallaty with best possible pc. specs

when the development of hardware in 10-15 years reachs a realtime realworld render possibilitie, than with every more progress they’ll add features and possibilities of modifiers, complex and layered materials and so on…
i think in about 20 years you’ll have the possibilities to take max files as realtime render engine models


This is a really great thread (took two afternoons to read through, lol), and has helped me increase my knowledge substantially regarding subD modeling practices. Many thanks to all who are involved!


Now for a question. Can’t remember what page exactly, but at one point I think setting creases were mentioned as being bad. Just curious if this applies only to organic modeling, or all types of subD modeling. For example, what if I have an object that probably won’t be animated, and even if it was it is never going to need to deform? My current WIP model is semi-organic in nature, and instead of using chamfers and what not to create hard edges, I made use of the crease setting. The results look good in areas that blend between inorganic/organic, and it also meant that the complexity of my model didn’t have to get out of hand (ok, so I’m slightly lazy too hehe). I’m curious as to what kind of comments this will receive and whether what I am doing is considered “wrong” at all, and why. Thanks in advance! :thumbsup: