PDA

View Full Version : Asset efficiency (modeling) and quality



hrgiger
05-16-2011, 12:54 PM
I was wanting to start a thread/discussion about modeling and efficiency in modeling and just how important it is to people. For instance, I just picked up a 3D magazine at a bookstore and it came with a resource CD. I was taking a look at some of the free models that came on the CD and it was interesting to see that they had a chair model that came in over 100K polygons. This struck me as interesting because I've always tried to be efficient when modeling and keep my polygon count as low as possible but balancing it against the final quality of the mesh. I feel like if I hit 100,000 polys while modeling something like a chair, that maybe that I'd have to see if I could reduce that somehow. Of course, the chair had a lot of ornate detail in it, so the count may have been justified.

Generally, I like modeling with subpatches/subdivision surfaces because if I see polygonal faceting in the render, I can control the smoothness of the model by up the subpatch resolution in Layout.

For instance, in the attached picture is a recent piece of furntiure that I modeled. It comes in at around 11,600 polygons. Which is fine in Lightwave, no problem. But what if I wanted to give this to another user of another software in a .obj file? Freezing the mesh at level 3 increases the polygon cout to around 104,000 polygons. I could of course reduce it to level 2 which takes it down to around 46,000 polygons but I don't like the faceting on the rounder areas. What good ways are there to decrease localized geometry without affecting the smoothing of the object? What kind of polygon errors, oGL or otherwise can this create when you export it as an object? For instance, if you look at the second picture which is the subpatch object frozen at level 3, is there a reason I shouldn't just turn a lot of the flat top of the dresser into a single polygon to reduce polygon count? I know that sometimes there are issues moving meshes from one program to the other and how they handle different types of geometry.

And that's what I'm looking at. How do I create an asset that will be most useful for others? Would it be best to offer both a object that can be used as a subpatch/subdivision surface model and a hard surface polygonal object? Or do people generally prefer to use models that they don't have to subdivide?

Would love to get some feedback or suggestions on this-

BigHache
05-16-2011, 01:43 PM
That's a great question. I think I would like to subdivide as needed from a lower poly asset, because at that point it's easier to increase the poly count than it would be to decrease it in a frozen mesh. But my 3D production environment is just me, so maybe that's backwards thinking?

WilliamVaughan
05-16-2011, 02:34 PM
These days I'm sure most artists would prefer a SubD mesh no matter what software they are using.

probiner
05-16-2011, 03:24 PM
How do I create an asset that will be most useful for others? Would it be best to offer both a object that can be used as a subpatch/subdivision surface model and a hard surface polygonal object? Or do people generally prefer to use models that they don't have to subdivide? As for this matter i have no clue.

As for the rest... I posted this thread (http://www.newtek.com/forums/showthread.php?t=117211) before where i propose a method to freeze SubD objects with more control over the tessellation. In your furniture for example the drawers would only have vertical tessellation and none horizontal tessellation (adds nothing) and the flat areas in the top and side would be n-gons.

Right now as it is implemented in LW, Catmull Clark Edge Weight is quite overlooked. But it adds great control of wires and in some situations a more reasonable/accurate shape, while allowing to have considerable sized N-gons sitting in flat areas and looking good. This will free the model from having to have all quads/tris in such areas, that otherwise, increase the polycount with subdivision, without any profit. (example: the top of your furniture could be simply, a single big flat N-gon, smoothing issues aside).
And this would have to have a correct UV Interpolation implementation as well.

Cheers

Danner
05-16-2011, 04:13 PM
I am probably too strict in making all my geometry as optimized as possible. I love under 4 minute render times I guess. I would run the frozen piece of furniture through the free plug in "PLG simplify mesh" and it would look the same at a fraction of the poly count. I'd would keep a subpatch version in case I wanted to tweak something.

If I had a dime for every piece of high detail object I made that didn't even make it to the final shot or was only a dozen pixels in the final render..

prometheus
05-17-2011, 10:37 AM
If you can...simply provide both the subpatch, and an optimized freezed version.

You canīt load in subpatches without freezing in some of the new render specific software, like octane and keyshot.

some people might be able to do the conversions themself or use conversion tools like deep exploration etc, but in case they canīt..why not provide both?

Michael

hrgiger
05-17-2011, 11:33 AM
If you can...simply provide both the subpatch, and an optimized freezed version.

You canīt load in subpatches without freezing in some of the new render specific software, like octane and keyshot.

some people might be able to do the conversions themself or use conversion tools like deep exploration etc, but in case they canīt..why not provide both?

Michael

I guess what I'm wondering is, is it sufficient to provide a mesh that is subpatch ready. In other words, a subpatch suitable model exported with subdivision turned off so that whomever uses it, can subdivide in in their program. I assume most subdivisions (differences between subpatch and CC's taken into account) work failry similarly among packages.

But yes, I thought maybe providing both would probably be optimal.

Nicolas Jordan
05-17-2011, 11:36 AM
When I model furniture for a rendering I usually concentrate on quality and speed and I won't normally spend extra time optimizing a mesh just for the sake of it unless it makes modeling it easier in some way.

hrgiger
05-17-2011, 11:40 AM
When I model furniture for a rendering I usually concentrate on quality and speed and I won't normally spend extra time optimizing a mesh just for the sake of it unless it makes modeling it easier in some way.

I completely agree for assets I use myself. I guess in this context I'm thinking of assets that I would give or sell to other users like on turbosquid or something.

Nicolas Jordan
05-17-2011, 11:56 AM
I completely agree for assets I use myself. I guess in this context I'm thinking of assets that I would give or sell to other users like on turbosquid or something.

In that case I would good to make the model in such a way that it could be used for a wide range of things It would probably make the most sense to use subdivision surfaces would probably be the way to go for most objects that way artists can adjust the density of the mesh in whatever program they are using.

probiner
05-17-2011, 01:52 PM
Also providing the models frozen would have the following benefits i guess:
- You control the render quality of the mesh, so if the software doesn't have subdivision or makes something different (edge weight, point weight, Doo-Sabin and Catmull-Clark do look different, open edges, etc) what you provide is unchanged and you have control of it's quality.
- Allows you to use UV interpolation knowing that it will match in other softwares (which doesnt' happen right now with LW, being linear still the safest way, but linear is not great, unless, it's from freezing SubD using interpolation, or the UV is made taking SubD deformation into account.)

Cheers

Nicolas Jordan
05-17-2011, 02:10 PM
Also providing the models frozen would have the following benefits i guess:
- You control the render quality of the mesh, so if the software doesn't have subdivision or makes something different (edge weight, Doo-Sabin and Catmull-Clark do look different, open edges, etc) what you provide is unchanged and you have control of it's quality.
- Allows you to use UV interpolation knowing that it will match in other softwares (which doesnt' happen right now with LW, being linear still the safest way, but linear is not great, unless, it's from freezing SubD using interpolation, or the UV is made taking SubD deformation into account.)

Cheers

You bring up some very good points.

cg_mike
05-17-2011, 10:50 PM
I guess in this context I'm thinking of assets that I would give or sell to other users like on turbosquid or something.

Most Turbosquid vendors offer multiple download formats. If those formats can be uploaded by the artist then you could offer, say, a sub-d LWO file, a frozen OBJ file, etc. and then reference the details of each in the description.

Titus
05-18-2011, 10:43 AM
When I model furniture for a rendering I usually concentrate on quality and speed and I won't normally spend extra time optimizing a mesh just for the sake of it unless it makes modeling it easier in some way.

At the same time optimizing models help speed your renders. I think there's no one way to do it correctly because every project is different, sometimes you need accuracy, sometimes you need lightweight geometry.

I'm with William; provide the basic cage, every modern program has subdivs.

prometheus
05-18-2011, 11:17 AM
At the same time optimizing models help speed your renders. I think there's no one way to do it correctly because every project is different, sometimes you need accuracy, sometimes you need lightweight geometry.

I'm with William; provide the basic cage, every modern program has subdivs.

I just want to be obnoxius here...today:)
no! not every modern program does have subdivs.

If you were to entirely work with octane or keyshot, then your stuck..but then again that would be very very stupid, since there are free programs handling that anyway.:devil:

Michael