clothing face count

[Grid Strom]

I am looking for help in regards to good acceptable face counts for clothing items. I have looked on the Good Building Practices wiki but I can not find any thing about this there.

I need to know what would be an acceptable face count for jackets and what would be the maximum one should go to.

Currently I am experimenting, creating a detailed jacket that would end up being around 5000 faces with zippers and pockets a collar and sculpted creases. Is this to much or is it an acceptable standard?

[Chosen Few]

As always, the answer is use only as many polygons as are necessary to convincingly portray the shapes you're attempting to represent, no more.  There can be no "one size fits all" answer.

5000 is a nice round number, but without knowing what exactly you're trying to make, it's impossible to say whether it's high or low.  If you can make the same jacket with less, than it's high.  If you can't make it from that many, then it's low.  If it's the least amount you need to get the job done, then it's just right. 

Keep in mind, whatever number you arrive at for your first draft, chances are excellent that you can lower it significantly upon doing it a second time.  There are almost always more efficient ways to solve geometry problems that don't reveal themselves until after the model has been made the first time.

[Rahkis Andel]

My personal (Blender) workflow:

I create a low poly base mesh of my concept that is as bare-bones as possible while still having all of the proper edge flows to get density where it's needed. I don't care how closely it fits to my concept -- I use as few polys as I can get away with.

Then I do multiresolution sculpting to make it match my concept. Once I have a polished high-res sculpt, i kick it back to the second subdivision level (or whatever level most efficiently adheres to the main shapes) and apply the multires modifier.

Then I manually remove edge loops until the mesh is as sparse as possible while still retaining it's shape. This time I really care that the low poly model stays as accurate as physically possible to the high poly. This will serve as my final game-ready model.

This is a great method for generating good high poly to low poly normal maps, which I presume you will some day want to do when the new material system comes out. Heck, you might want to do it now just for the benefit of those who are using the test viewer.

If you wanted to go hard-core low-poly, you could triangulate your mesh and start collapsing edges in areas that don't need the complexity, but realize that will make your mesh unreasonably difficult to modify later. I recommend keeping an all quad version at all times. Also realize that you are not going to see significant gains from sparing a few polygons here and there (in my opinion).

[Nalates Urriah]

Both Chosen's and Rahkis' answers are good. 

I'll add that Materials are coming and that will enable normal maps. If you look at what professinal modelers are doing, you'll see they do a higher poly model and bake a normal map from it to put on a much lower poly final model.

Consider this in you design. In the next month, may be 2, we will see materials rollout. The server side is already done. The viewer side is completing now and there are versions of the Materials Project Viewer rolling out every week. Once that merges with the Developement viewer we will only be a couple of weeks away. Overall I am guessing 4 to 8 weeks.

[Codewarrior Congrejo]

It really depends.
As you heard already the lower the better for everyone. (less renderimpact, and ressource costs)

In general we can make a full avatar with just 6 K polygons. And the upper body of this avatar could have on its own around 1300 polys i.e.  So if a jacket alone takes already 5 K you can consider this from certain points of view of course a lot or overboard.

But of course if it has a lot detail, the amount of polygons will differ from an undressed upper body and thus require a bit more polygons.

But again speaking of detail, here you will have to decide 'how much' detail in actual gemoetry is needed and how much of it can already be achieved by textures and certain bakes.

F.e. the Zipper, you really don't need to shape out every bulp and indent in the zipper. A straight polygon faceloop with possibly sideparts and a back side is mostly already enough. And if you don't shape out every zip-piece in the zipper this can already save you 1 K polygons just on this 'detail' of your jacket. 

Consider that mostly people will not even be closely zoomed enough on the jacket another user is wearing to optically differenciate if the Indents are just shadows in the texture or actual geometry on such small details as the zipper parts.

Another example: the pockets: they surely don't need a 'backside or inside" geometry and can just be sunken into the jacket. If they are not intended to hang half open and allow a look inside there is no need to actually create polygons for their inside parts. (same goes for saving all polygons on the parts inside of the jacket, mostly 1 or 2 faceloops inwards are enough to cover the area one could view from certain view angles)

Folds, Wrinkles: also these don't need to be actual geometry, most of them can simply be done by the texture. and only the ones that 'really' stand out or are sunken deep in, should be actually formed out.

Also keep the amount of edgeloops and faces low in areas that are not bend or influenced by animation. You only need 3-4 edgeloops close by each other in spots like the knees, shoulders, ellebows and the belly middle. (etc)

Having a continously even flow in the density of faces does 'not' need to lead to the fact that because  i.e. the creases have a lot, thus the other areas need these too. It should just be a generally 'evenly' spreading of the faces. But they can surely vary in different areas.

Once the jacket is done, and you feel you did it with a reasonable amount of polygons, you should / or can as Rahkis already described start optimizing, by removing whole faceloops, and centering the remaining 2 edgeloops on the sides of the removed one, to keep the same visual shape. And remove or collaps all faces, merge vertices and much more,  where you figure they are unneeded and just a result of the subdivision modifier multiplicating them on these spots.

And as Nalates already stated too, with the upcoming normals and specular map features you will be able to break it down even more by having folds and certain details completely done by texture mapping.

[Sienia Trevellion]

I'm no expert, I'm always coming here for help but this tutorial really helped me with what she is saying. Creating a high poly for textures and then using the low poly model. It's a great one.

 

http://cgcookie.com/blender/cgc-series/low-poly-game-asset-creation-fire-hydrant-blender-unity-3d/

[Grid Strom]

Thanks for the reply’s

I have remade my base topology so that a basic turtleneck top has around 2400 faces instead of the 3900 with no extras I had before. I want to create efficient, but also good looking clothing topology. I was able to bake a normal map from a much higher resolution model that I sculpted in the creases. I uploaded this and tested it on the new materials viewer it looks ok but I think there is some thing lacking, maybe the defuse texture still requires a little shading here and there? As with my normal mapped top it looks very cartoonish.

Currently I have 18 faces going around the arms
36 going around the waist
24 going around the neck

This seems to be a nice base to work with it. I have been looking around SL in wireframe mode to see how other creators are doing there clothing topology and comparing it to mine, I see that I used a lot less with my lower topology attempt.

I must admit this is not an easy thing to work out because I want to be using the same amount of faces as every one else but then I want to be efficient.

Last question: Do people really care about efficient topology?

[Kwakkelde Kwak]

---

Grid Strom wrote:

I uploaded this and tested it on the new materials viewer it looks ok but I think there is some thing lacking, maybe the defuse texture still requires a little shading here and there? As with my normal mapped top it looks very cartoonish.

---

You could add baked in shadow and occlusion effects to the diffuse maps. That will certainly make a difference. You can use your high poly model for that as well.

 

 

---

Grid Strom wrote:

 

Last question: Do people really care about efficient topology?

---

I'm afraid most people would say "no" if you'd ask them, they'd rather have the "extra detail". Those same people do care about lag though. They also care about weird deformations of an object when animated. If it's not a worn object they also care about landimpact and objects breaking up from a short distance. Some might not like the fact some objects are difficult to texture.

Maybe you're asking the wrong question?

My personal opinion on "detail" in SL is not the same as it would be for a 3d game. In SL every avatar is a primary character for the person using it and secondary to all others. An object is a key object for anyone using it and scenery for anyone who doesn't. So I would say all SL content should be between primary and secondary, call it one-and-a-halfy.