View Full Version : Proxy models - Wrap deformers or Rendertime SubD
09 September 2007, 01:11 PM
Hi I got a question to people who know about current Studio trends.
I would like to know that when animating a Character what is the preferred method that is used by Studios switching between the Low res character that is animated and previewed by animators to the Hi-res model that is rendered for the final output.
I know that there is a method when the Pipeline is like that there is a Maya Subd as the Hi-res, and a lowpoly that is deforming the Subd via a Wrap Deformer. I know that this has a few issues, such as the wrap deformer has limits, and may need to be cut up. Also Not sure if this kind of setup is accurate enough these days for the hi-res deformations.
Or there is the other way, that There is a higher-res Poly mesh, that is SubD-d at Rendertime, and that is Skinned to the same bones that control the lowres Poly?
Which one is better/more accurate and the preferred way by studios these days?
09 September 2007, 04:07 PM
One common pipeline would be using Subdivision/Polygon Approximation at Rendertime via Displacement Mapping. The low res base model is what is being rigged and and the detailing is done externally in i.e. Zbrush or Mudbox.
In-Between steps are feasible aswell, i.e. a low res mesh that is rigged and skinned, wrap deformer on higher res mesh with additional fancyness which then is approximated and detailed at rendertime only.
09 September 2007, 04:17 PM
Thanks for the reply, my main question was really about if the "wrap deformer" method is outdated by now or not ? To me Wraps seem to be a bit more of a guesswork to get the deformations correct and accurate? Or maybe im totally wrong and its still used in production ?
I mean yes it is a whole lot more feasable just to have Rendertime Subd(especially with Renderman based pipelines).
09 September 2007, 04:39 PM
I am not particularly familiar with the wrap deformer as a low to hi res bridge but I am aware that people still use it frequently. I am however not working for one of the big studios so maybe this is not as common as I think.
09 September 2007, 11:43 PM
they were invented to skin nurbs patch mesh characters, which can otherwise be a nightmare to weight in such a way that the patches maintain tangency during deformation..
so if you are using smoothed polygons or subDs for your final rendered mesh its an utter waste of time.
and as far as i know even the big studios have abandoned nurbs character modelling.
09 September 2007, 09:11 AM
Not sure if I agree, the same goes for polygon meshes, particularly if you have many blendshapes that only really affect certain areas such as the head (but not the body). In my experience splitting the low res mesh up and using the parts to deform the full (one piece) mesh makes a lot of sense.
Also, whilst I agree that nurbs character modeling is somewhat of a rarity nowadays in production pipelines, I am not sure where the sudden talk about NURBS comes from in the first place as there was no mention of it anywhere in the previous posts? :)
09 September 2007, 10:02 AM
I was more interested in a quality comparision, like if Wrap deformers are accurate enough to deform a hi-res mesh compared to a Rendertime SubD?
09 September 2007, 10:05 AM
two very different things. one just smoothes and/or displaces the model, the other one controls deformations based on weighting etc.
You (can) use both. I do.
09 September 2007, 02:56 PM
Yes thats what i meant, the Wrap deformer would be controlling a SubD- that has Displacement as well.
But what i was asking is that if the wrap deformer can correctly translate the deformations from the lowres to the hi-res(that could be displaced after or whatever)
Where as a Rendertime SubD is just simply subdivided so i would assume that the deformations are 100% translated.
09 September 2007, 02:59 PM
Inherently yes, but it entirely depends on how well it is rigged.
IMO the answer to your question is skill, not a debate about a particular tool.
Try it out!
09 September 2007, 10:00 PM
it probably CAN but it wont by default :)
i still dont see how you need wraps on a polygon mesh ever, if i want to rig my head separate from a seamless body i just make it separately, make a copy and define that as a blend shape target, and then attatch it with construction history and merge all the verts. using the original unmerged one as a blend shape target i can then add as many blendshapes as i want to that one at any time, and the final seamless mesh will update properly. then you can just smooth your final merged mesh, or apply a rendertime subD approximation if you are into that sort of thing :) easy! and your rig will be light as hell, no wrap popping out of place and bogging stuff down by having connection to a super high res mesh that bloats up your file, etc etc.
i am actually doing this right this second, on a very legitimate commercial job :) i'm just waiting for a test render while i check my messages. so i know this workflow i outline is not BS.
the reason i mentioned NURBS, if you bothered to read my post, is that wrap deformers were invented as a tool to compensate for the difficulty of weighting nurbs patch meshes. thats the only reason its included in maya at all, really. the fact that you can use it for other stuff is because maya is cleverly designed to be flexible but there was an original specific reason for the wrap, and its nothing to do with polygons :)
so a better question would be why the original poster mentioned wraps but did not meniton nurbs IMO ! the two are inseperably liniked, like superman and clark kent.
09 September 2007, 09:50 AM
The reason I mentioned is that if you happen to use Renderman For Maya, and displacements from Zbrush then its very difficult to get them work together in a Rendertime SubD, but they work relatively easily in a Maya SubD. And for a Maya SubD you will need a wrap deformer.
Its not impossible to get them line up, but its difficult, so thats why I wanted to know if Wrap deformers are accurate enough, because if they are then maybe its a viable solution, but if they arent, then its worthwhile to find the right settings to get displacements working in a Rendertime SubD.
I just thought you wont need to know about this, but there you have it, its probably more than what you bargained for :D
09 September 2007, 07:55 PM
well well. yeah fair enough, i guess i would say since i have no personal experience with your renderman/maya/zbrush pipeline, i cant begin to weigh the difficulty of that task vs. wraps. all i know from experience is using wraps, and i know i dont use them if i can possibly avoid it :) it is very likely that wraps will cause other headaches which may or may not outweigh the issues with renderman, roll the dice...
that being said i still dont see how this necessitates wraps. if the problem you have is simply driving geometry thats subdivided 30 levels for the displacements to work sweet from a rig without it getting insanely slow, theres many many ways to do this beside wraps. if you simply rig your level zero subdivision before increasing the smooth, thats the easiest :)
if you are playing with high end stuff like renderman, it should be possible for you or someone in your production to write a script that bumps up your maya subD when u launch the render, if you really want to keep the scene light.. or else if you arent THAT legit but not totally a wanker, you can make an attribute in your rig and hook it up via the connection editor to the appropriate input on your sudB to control the divisions. this is also nice cause its easy to define custom values for different parts of the mesh, i.e. i commonly have an over ride on characters faces so i can smooth them more than the body for close up shots..
this is what i do cuz i'm not a scripter, it becomes part of the routine to turn the value up before render but i realize in a big shop they'd want something more idiot proof, as it is always possible to simply forget to smooth it, or input the wrong settings.. not that i ever do this of course.. ;)
09 September 2007, 09:49 AM
Well you really asked for it now:), I have to give full explanation, So the problem is not with the geometry itself, its about the UV coordinates and therefore the displacement maps lining up to the SubD geometry when you Render it.
The problem originates from that when you Subdivide a geometry in Zbrush, it interpolates the UV coordinates differently as Renderman expects them. So consequently the displacement maps generated will not be correct either, and as a result you will get not only bad UV seam issues, but the whole displacement will be moved off target, as all the internal UV coordinates are interpolated badly as well.
However if you use Renderman with a Maya SubD, then everything will line up correctly just out of the box.
Recently I figured out a pipeline to get around this problem, to be able to use Rendertime SubDs, I even wrote a guide over Zbrushcentral (http://www.zbrushcentral.com/zbc/showthread.php?t=50873).
PS: Renderman being Hi-end? Not really(pricewise), these days Renderman renderers don't cost more than any other Renderer. There is very good Renderman Renderer called 3Delight, you actually get the first commercial license totally free. Even Pixar sells Renderman for Maya for $990.
09 September 2007, 09:49 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.