Increasing / Decreasing detail

[nervous_twitch]

Nope. Not in Wings.

Sorry, I'd like this too.

[wasamonkey]

nborg you are refering to real sds
only one commercial app has this
maya
pixar also has real sds and developed the technology and seam to own the paten on it as they are the ones that sold maya the rights to use it in maya4.5

[Nborg]

Oh, I didn't know that. So.. what kind of subd technique do you guys use in Wings? Or in other words, how do you work around the problemmo?

[MasonDoran]

its nothing more then just smoothing the surface....ie: subdividing....you are always looking at polys...Mayas subDs allow you to subdivide in regions without actually subdividing the poly mesh...you can get an extremely fine level of detail....but not without a cost(which is why most modellers prefer the fake subds of the other apps cuz they work like polys)

[MasonDoran]

the art of modelling is learning how to model on low poly cage....knowing in advance what the smooth version will look like....what this does to the modelling process is people will often exaggerate the angles and volume of the mesh to compensate for when it is smoothed.....


i have actually started to prefer the raw model over the smoothed version...there is something special and original in these rough shapes....they already capture the element of the character(and artist)...kinda like concept sketches....

[Nborg]

2byts: I agree somewhat on that. I really enjoy looking at a well modelled model before smooth is applied. Bay Raitts raw models are awesome!

So when modeling in wings, one just models regular and then smooths when done? That's at least how I do it.

[somlor]

Originally posted by wasamonkey
nborg you are refering to real sds
only one commercial app has this
maya
pixar also has real sds and developed the technology and seam to own the paten on it as they are the ones that sold maya the rights to use it in maya4.5
I believe XSI also uses true SDS algorithms and even allows you to choose between Catmull and Doo Sabin.

[s]

[wasamonkey]

solmor ah hadnt heard that
will have to check into it

[MrMunkily]

yes that is the case for XSI.

as far as choosing catmull/sabin is concerned. I don't know if they are "real" SDS though

[dogz]

if this helps,, you could low poly model and go to-file-render-and then click that little box next to the open gl shift r menu and you get a pop up menu.. towards the bottom it lets you control the subdivision level of the model when it renders..but wont affect the geometry you are wotking on ..i find it very usefull for checking how smooth my model will be ..but turn down the quality of the render to draft.. so it renders faster..its the box above the subdivision menu..hope this helps..

[vizion]

hey that's an awesome tip dogz. Thanx.

[BazC]

If 0.98.08 works for you you could have one window with a low poly cage to work on and another window with a smooth preview view open at the same time. It would save stopping and checking smooth preview or rendering every few minutes!

Baz

[Mauritius]

pixar also has real sds and developed the technology and seam to own the paten on it as they are the ones that sold maya the rights to use it in maya4.5


This is not quite correct.

Maya uses hierarchical Catmull-Clark sds with creases based on degree 3 NURBS cv multiplicity (only two crease levels).
Pixar doesn't own any patent on either of these three techniques (C-C sds, hierarchical sds or NURBS cv multiplicity).
Pixar does own a patent on arbitrary sharp creases on C-C sds (which Maya doesn't support), as found in Avid's XSI (who likely licensed this from Pixar).
These so called 'extensions' to the original C-C algorithm, which also include 'holes' and interpolated surface boundaries, are described in a paper by Tony deRose et al. For this reason they are also referred to by geeks as the 'deRose extensions'.:cool:

You can use C-C sds with arbitrary sharp creases in Maya through e.g. the MTOR plug-in from Pixar, tag Maya polygon meshes and their edges/vertices, then render these through PRMan, 3Delight or Pixie as sds (all three renderers support sds including the deRose extensions).

Cheers,

.mm

[wasamonkey]

This is not quite correct.

Maya uses hierarchical Catmull-Clark sds with creases based on degree 3 NURBS cv multiplicity (only two crease levels).
Pixar doesn't own any patent on either of these three techniques (C-C sds, hierarchical sds or NURBS cv multiplicity).


would you tell that to some of these ppl on cgtalk :)
they are the ones that gave me this information and said that only maya and pixar have real sds :)

[Mauritius]

Well, to be honest only very few renderers have real sds from a geek's pov. Maya is not among them.

Here are two interesting mails on the subject from Larry Griz to the Highend3d rendering theory list a bit over a year ago:

http://www.highend3d.com/render/archive/sp.3d?mail_id=1523
http://www.highend3d.com/render/archive/sp.3d?mail_id=1521

In particular the second one propses a 'Turing test' to determine whether your package/renderer of choice does support sds or not ...

.mm

[somlor]

So "true" SDS from this perspective, is actually a function of the renderer which calculates the level of detail on the fly, based on proximity to the camera, user settings, and perhaps even adaptively as needed, for example, if a polygonal surface is curving it will know to refine this area more. (?)

If my assumption is correct, does that mean that any package capable of working with polygons and outputting to an SDS capable RMan renderer qualifies as a "real" SubD modeler?

I seem to remember you mentioning before that there is more to the story; that the typical subD toolset itself is lacking.

[s]

[Mauritius]

Indeed there is -- read the last paragraph of the first mail I linked to on poly- vs NURB- vs sds modeling and you see that there isn't a single sds modeler available today if you're honest -- only poly modelers.
The point here is not that the display of the limit surface in the modeler is of approximative nature. The point is that no package allows you to work with the limit surface itself -- e.g. project a curve on it or push/pull the surface directly, rather than the control cage.
Another example: A "real" sds modeler would perform boolean operations with the resp. limit surfaces, whereas today's so called "sds" modelers perfrom boolean operations on the control cages ...

.mm

[wasamonkey]

Mauritius: Thanks for the read
great stuff

I remember reading something about what a real sds modeler would be able to do

would mayas handling be the closest commercial app to having true sds modeling?

[Mauritius]

No, not closer than aynone else.

.mm

[Thalaxis]

Originally posted by somlor
I believe XSI also uses true SDS algorithms and even allows you to choose between Catmull and Doo Sabin.
[s]

Cinema4D and LightWave and RealSoft also have "real" SDS algorithms.

[Mauritius]

For Cinema4D I can't really tell (see below). But for Realsoft3D I can: The answer is no.
RS3D uses a Newton solver to calculate a direct intersection with a NURB patch (sds get converted to NURB patches before rendering). However, the precision of this algorithm, which is recursive and must be terminated at some point, is selected by the user, not the renderer. That is, as you zoom or dolly into the geometry, silhoeutte and shading artifacts show up that are similiar to those seen in other renderers that do not take object size in screen space into account to determine tesselation density.
I heard that you have to specify a subdivision steps value for Cinema 4D before rendering also. If this is the case, it fails the suggested 'Gritz test' also.

Currently, only micropolygon renderers that can dice depending on object size in screen space pass the 'Gritz test' .
This includes most RMan compliant renderers and MRay. Houdini's VMantra is also a candidate and, I guess, Jig.
MRay btw. only dices into micropolygons of the size commonly found in RMan renderers, if you set 'view length' tesselation to 1 or smaller.

Cheers,

.mm

[CGTalk Moderation]

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.