Why Subdivs surfaces? Why people still use them?
Ok i guys i am currently learning maya in my free time and i love cg stuffs when i am learning modeling i mostly use polygons (nurbs 10%) but the mighty "INTERNET" says Subdivs are the great u can do organic modeling very easily.What's wrong with polygons? they give us a great advantage of their edge,vertext,faces editing right? and also u can use smooth node or press 3 in maya to Smooth it and do real time modeling if u want also we get the same result as subdivs surfaces. so why people use their time to convert to subdivs surfaces???
i just asked this question to clear out a problem of mine but not to insult people's workflow if it sounds like sorry i hope you give me a good explanation.
i just asked this question to clear out a problem of mine but not to insult people's workflow if it sounds like sorry i hope you give me a good explanation.
Replies
SDS is only a special case there, nornally subdivisian surface means polygon modelling with some sort of meshsmooth applied
Check out this epic thread: http://www.polycount.com/forum/showthread.php?t=56014
i think you did not understand the posts before
@op: actually nope, almost nobody does
Could be the case! Fixed post for clarity.
well I mean..
It's when you add the word surface that it becomes a technical term! But naming ends up quite silly anyway.
Diffuse/Albedo etc.
subd modelling - modelling polygonal stuff with the intention to use a subdivision or smooth modifier on it
mostly to create a nice and smooth highpoly for your bakes (at least in game modelling)
subdivision surface - maya only (?) type of surfaces, much like nurbs surfaces, but with region control. i havent really used them, only for conversion to import into deltagen, so i am not too sure what benefits they have over nurbs
nurbs modeling - surfaces from curves(splines)
Now of course at the end of the day it's not a huge deal, similarly to the way some artists use "shader" interchangeably with "material" ... it's just a bit of a shame that this kind of liberal use of technical terms ends up causing confusion.
might be true, but people who use this on a daily basis usually know whats meant, and thats the important thing
it would look even more silly if i were to define it with something like "turbo smooth modelling", because nobody uses this term
so.. yes.. its stupid, cant do anything about that
if anyone has a better explanation, give it a shot
I dunno if it ever could, or if any developers ever actually did.
I guess we're unlikely to ever see NURBS appear in engines again given decades of optimisation towards polygons, but it seems quite likely that as your crazy complicated Unreal 4 guns start surfacing, some bright spark is going to realise how much quicker they could be working with a product CAD app.
depends on the software though
autoCAD is pretty destructive while in programs like catia or inventor you are able to change things up
This looks promising for CAD workflows. I could see building a weapon in it. It allows non-destructive workflows. It saves every step you make.
100$ per month. Wowzers. Those rental models could add up fast.
One of these days when I'm not super busy or lazy (read: probably never), I'll have a crack at knocking out gun in Solidworks and see about getting it into U4.
It's a technique of modelling, as simple as that. Long time ago, in max it was called NURMS modelling ¬¬
Check out some modo videos in youtube, and you will understand why subdivision modelling is so important. It will be the future for games.
Try to texture map something NURBS and you'll immediately understand. Hard surface models in particular are really evil, as you can't use the implicit parametric UV because of all the fillet and trim and other surfaces. A complex model can consist of hundreds of these and some of them are just tiny strips and such. So the usual method was many many projections, each with its masks, I think. This is one of the strange legacy things in Maya where every texture you apply on something immediately gets a projection node in front of it in Hypershade - on polys nowadays it's just a 2D projection but there's also a 3D type. So shading networks for a complex NURBS model were HUGE.
The other option is to convert to polygons and add UVs, but then you'd lose all the advantages and also end up with a pretty messy model. Not to mention how hard it is to work with a multi-million triangle mesh in the UV editor.
One thing is sure though, you can never get that level of quality in your surfaces with subdivs, which is especially obvious on highly reflective objects, like a car.
http://www.cgfeedback.com/cgfeedback/showthread.php?t=6610
Also, I've never used Alias Studio or some other advanced NURBS modeler, but I can tell you that the tools in Maya have always been quite fragile. Even something simple like rims for a car took a LOT of time with all the intermediate steps; and every now and then some operation failed and created holes and such and the only way to fix it was to go back several steps and redo it all.
Personally I've also found that hard surface modeling with NURBS wasn't actually faster or more efficient at all. Setting up a complex intersection with trims and fillets and blends and lofts was a LOT of work... Oh and you had to learn a lot more theory just to get started, things like continuity and tangency and such. Granted, I've only spent a little time with this, mostly in our Maya training back in 2001 or so.
Creatures were at least somewhat simpler to texture because they were made with patch modeling - each patch had its own set of textures. Although most creatures required up to hundreds of patches, and thus as many textures for each channel. But at least in movie VFX it was a requirement anyway, in order to get good texel density (which is why UDIM was created for subdivs).
However patches tended to crack open during deformations, as they weren't really merged, only held together by code (which also made them extra slow). Oh and displacements tended to create the same kind of problems, and these were usually only fixed in comp with retouch and roto work.
Oh and it's very very hard to model organic stuff with NURBS patches, you have to fight to combine the complex tech and the artistic goals - so most studios started with scanned maquettes instead. I think the most complex character ever was Draco in Dragonheart; ILM nicknamed him "Tippet's revenge" (Phil sculpted the dragon) as it took several months for 2 or 3 guys to build the model and it was so heavy that they couldn't even load it into Alias (back then they only had 96MB RAM though
If you're more interested about the creature stuff, there used to be a SIGGRAPH paper in PDF about Disney's work on Dinosaur where they talk a lot about the issues they've encountered. Fascinating stuff, especially in light of how easy it is today to build a creature at that level of quality - making their accomplishment even more impressive.
http://renderman.pixar.com/view/rendering-issues-on-dinosaur
Also interesting are the shadow map issues, BTW - something game programmers have managed to solve pretty well in the past years.
NURBS still has its uses in CG work though, but mostly for complex effects related stuff, where the built in parametrization can be used for all kinds of procedural magic. But it was such a huge relief when Pixar introduced subdivs to PRMan, everyone transitioned immediately - Toy Story 2 converted all the old characters, and Weta switched between LOTR 1 and 2, for example.
So NURBS is more suited for industrial work; but even there, for example in the car industry, all the design work is done with polys nowadays, and the Alias guys only take over when everything is locked down. It's fascinating stuff though - one of my guys has left a few years ago to do this kind of work, for example he worked on the dashboard of the new Mercedes S coupe and he always tells awesome stories.
TLDR: The entertainment industry has abandoned NURBS for pretty good reasons, more then a decade ago.
Also, it's good to question existing beliefs and practices; but one should look into how they became standards as well
I'll admit I'm only dipping my toe into these waters at school, but I was really surprised by how quickly and efficiently models could be put together in Solidworks. It's quite unlike the needlessly cumbersome set up of Alias and has one of the better UIs I've seen in any 3d package.
As I see it, It would be a specific tool for a specific purpose in games - knocking out hard surface models to then be retopolgised in polygons. It's probably something that'd take a little development both in terms of skill and software to get the most out of it, but I could see definite potential there as the need for complex hard surface geo increases.
Some interesting reading there though. Thanks for taking the time to go through all that.
[ame]
This technique already exist in Maya,Max and XSI,don't know about Blender.
http://blenderartists.org/forum/showthread.php?329295-Bevel-shader
Well to be fair Maya's way of working with it's Mia_RoundCorners is really tricky too,but there are some really nice results with this technique using Maya (Courtesy of Fingus) and even far more better results with MODO in a very short amount of time (one hour for this gun,courtesy of Tor Frick) .
Actually,i don't even understand why these methods aren't supported enough because it's seems to be obvious that once you have a solid pipeline based on it and some additional features here and there it can be really robust and for an industry who's always looking for high quality work in a short amount of time,this shader from what I've seen is really strong,far more faster/stronger than Sub-D modeling in normal typical hard surface scenarios.
Also, will Zbrush become more prominent in 3D modeling pipelines with the new Zmodeler function included?
In the end of the day,there are many solutions that are far more better to create assets quicker,and using Sub-D in this case becomes more easier too since you only do it if it's necessary.
Warren did a [ame="
Is it easy to integrate in any pipeline ? not yet in the industry but i think sometimes you might want to share your assets and maybe the high-poly.If your studio isn't using MODO as the main modeling app i don't think you're athaurized to use it (wheighting),right ? or is there a balance ?
One possible approach is to use the meshsmooth/turbosmooth option in Max where it keeps hard edges between different smoothing groups. Example workflow:
- set up your smoothing groups on the base mesh
- add a smooth modifier that separates based on them
- add another smooth modifier that doesn't
You may have to tweak the subdiv iterations on the two modifiers, sometimes 1/2 works better, sometimes 2/1.
Can you elaborate on that? I'm genuinely curious because there's a 50/50 chance of my heading back to Modo this year. I thought opensubdiv made it so you can see edge weighting in most 3D apps even if you did it in a different app.
This video shows what I'm talking about:
https://vimeo.com/21625869
I take it back! I just exported a test mesh from MODO and pulled it into MAX 2015 ... there's some new modifier that MAX has called "Open Subdiv". It added it to the mesh automatically and there we are.
INTERESTING.
Interesting ... I didn't know this was even doable.