Quote - ...oh!  i see!  yes, that would pretty much kill it for creation for me, i think .  it's hard to say without understanding the mesh it makes better.  the thing is, wide skirts and trains with a topology like a cone can have really bad problems.  i learned that the hard way on my first piece.  i actually don't think it's posted here.  i messed up pretty thoroughly.  like mega bad.  i had to take it back into Blender and subdiv and smooth it.  now when i want (big) circular skirts and trains, i give them a grid topology rather than a radial one.  that isnt' to say that's the correct answer, just whew that first big wide circular train i made was just really awful. i just didn't think enough about the topology.

...

That's where I'm at right now. Skirts don't drape quite like I want them to. I mean, they don't look horrible, but just not as good as I think they could. Maybe I'll have to rethink what I'm doing there... The only downside is, it's not as easy to do what you did in Blender in Wings...lol. It can be done, but not as easily as you seem to have done. And there's no cloth sim in Wings, so I'll have to bounce back and forth from Wings to Poser, which really isn't that bad a deal...lol.

I wish I could wrap my head around Blender, but it just won't sink in...hehe.

Laurie

The blender plugin looks like it works like the Max garment maker.  I've had experience of that and it's pretty good fun :)

I know what you mean about using Cones for skirts, it can cause odd results.The type of skirt I'm doing dictates what I use.  If its a tightish shirt I will use a cylinder, merge it into V4 and go from there.  For a looser flowing skirt I use a cone, but it might as well be a flat disk with the angles I use :)

As to the poly placement you have reminded me of something I'd forgotten.  Max doesn't use regular polys at all in the garment maker, it makes a fairly random arrangement of quads to fill the space.  This gave a really nice flow to the cloth, I'm gonna to a skirt that way when I get home :) see how it comes out.

3D Coat is pretty alien as far as traditional polymodellers go, so I often forget to mention things that people wouldnt know if they didnt use it.  So appologies for the confusing descriptions :)

John,

oh, no, don't apologize.  i think you've done a very good job of describing something that works very differently than polygon modelers. 

in terms of the cone vs, cylinder: i do kind of the same thing.  so for a circular skirt i try to work with an actual circle.  it's kind of tricky my way because of how everything works.  right now i "sew" the top, then attach the skirt. i may find a better way to join the two.  for something that's more A line or straight hem skirt, i work with planes and just "sew" them like i did the shirt.  my latest skirt was gathered, so you can still use that method on a loose, flowing skirt.  but i suspect it will work much like a skirt actually made that way and not bell out in a circle.  on the other hand i have a sort of princess dress with a big long train that i made that way.

oh, and LaurieA, i totally understand about Blender.  i think comfort with modeling programs is really individual.  i'd think you could do the same thing in Wings?  so let me see...

I can do the same thing in Wings, it'll just take more time and effort. Not that I didn't want to learn Blender anyway, but for me being a new modeler that's never really modeled, Wings was the thing ;o). I guess I'll pony up the bucks and buy the manual for Blender. That might go a long way to helping me understand it...

Laurie

LaurieA - think you'll be much better than i am at it in a short time.  you've already made clothes a lot more complex than i know how to.

Also, in this image the dress is tri'd. Kaibach suggested it to me. It makes sense, since you want the verts to bend in any direction. And I did it [X] tri'd, not [/] tri'd. Still looks nice ;o).

Laurie

Don't know if this would work in all cases, but I got the cloth density pretty high .200 and the fold resistance pretty low 0.500. I'll have to do more experiments with different poses...lol.

Laurie

wow, cool!  i've been avoiding tris because someone advised against them in Poser.  i'd heard that they were good for cloth for exactly the reason you give, but then i read never use them in Poser.  this is very helpful information to have.

You can thank the eager mind of Kaibach for that one....lolol. I'd heard the same thing you heard and so I'd avoided them too...lol.

Laurie

Hi folks, just got a reply from Mark Bremmer-the BB of Carrara.  There is no mesh locking available.  However, beginning with Carrara 7, you can do all your fitting in the assembly room and work your way around the manikin-plus you can use the whole figure.  In Carrara 5 Pro, you have to work with hips, chest, thighs, shoulders etc separately in the vertex room and fit them together in the assembly room errrgh, errgh, triple errgh.

Just info for the Carrara model builders that are also Poser addicts (like me).

Laurie, what a neat looking dress, love it.  Jan

Well, when you think about tris for dynamics it makes sense. You wouldn't need them in a figure or a conformer, because those don't need to bend in EVERY direction. Dynamics potentially does ;o).

Laurie

Thanks Jan. Too bad your version of Carrara won't lock parts, but would be a good time to upgrade...lol ;o).

Laurie

Hi Laurie, LOL--waiting to get me hands on a free copy of Carrara 6 Pro-if it showed up in Creative Arts, just waiting for it to show up in 3D World.  I'm cheap-never been the first person to buy something new-when its new to me it's old for everyone else.  Okay I broke down and purchased v3 and v4 morph ++.  Besides, for me to download Carrara 7 pro, well, the sun will probably become a planetary nebula before the download finishes (4.5 billion years from now).  Jan

Oh it has not constrained groups at all, one nice thing about 3d coat is that you can model so close to the original that slippage becomes much less of an issue.

John

Quote - Just had a go at a quick skirt, what do ya think?

Oh it has not constrained groups at all, one nice thing about 3d coat is that you can model so close to the original that slippage becomes much less of an issue.

John

Hey John, do you find that it drapes nicer for you if you tri the model? Just asking folks to see what they think ;o). Seems to work better on my dress at least ;o).

Laurie

Quote -

Hey John, do you find that it drapes nicer for you if you tri the model? Just asking folks to see what they think ;o). Seems to work better on my dress at least ;o).

Laurie

Hiya Laurie :)

I think it does you know :) and I didnt make the quads ordered either, I started with big tris which 3d Coat turns into quads when you subdivide.   Then I used the triagulate mod in Hex to get the effect you were talking about :)  So I kinda took the ordered polys out to start with, then made even more flexible with your tri thingy :)

I use tris in all sorts of places that people tell me not to, mostly because I havn't really found a problem in doing so.  I am reliably told that Daz Studio really doesnt like tris, do if I were doing something for that I'd leave em out.

John.

I don't think we have to worry about dynamic props in D|S anyway. Not compatible ;o).

I started with my quaded dress and just tesselated it into triangles to get mine. So far, it's working...

Laurie

Quote - Just had a go at a quick skirt, what do ya think?

Oh it has not constrained groups at all, one nice thing about 3d coat is that you can model so close to the original that slippage becomes much less of an issue.

John

very cute! i'm curious, what is the UV map like?  i have very little experience dealing with UV mapping things that aren't flat.

just a comment: in my experience, regular topology is much more important than tris vs. quads.  a long thin tri is going to be a problem like a long thin quad would.  even if you want tris in your final mesh, the way you make it should give you as regular a topology as you can make.

and a quick question: how does this affect your polygon count?  can you get more flexibility with a lower resolution, do you need a higher poly count, or is about the same?

Attached Link: max garment maker

The UVMap is pretty flat, the skirt itself was mapped when it was flat, so has no real distortions in it.  The Waist part is unwrapped using the ABF method so it pretty flat.  It would be better if it were made with flat panels of course,  I will work on that in the future.

You are right, the polys should be of a more uniform size, it was just a quick test so I wasnt really thinking about that.

The polycount I suspect should be roughly the same, since I would not subdivide as far.  The las subdivide I would normally make would be replaced by a triangulation modifier, which is essentially a different way of subdividing.

Check out the link above, it shows some images of how the Max Garment modifier arranges the polys for a good cloth simulation.  As per Lauries suggestion it's all tris except they are arranged in a non linear manner, thats the effect I'm trying to get :)

John.

Fugazi1968 - oh!  the comment wasn't on the skirt.  it was on the whole tris vs. quads thing, and i was thinking about any lurkers here who haven't modeled dynamic clothes yet.  the mesh looks more than regular enough to work well.

just for discussion: have you done much using the opposite effect?  using topology to create structure and invoke folds where you want them to be? and have any of you done much work with pleats?  i have a vague idea for a process for making them, but i'm still kind of stumped.

Quote - Fugazi1968 - oh!  the comment wasn't on the skirt.  it was on the whole tris vs. quads thing, and i was thinking about any lurkers here who haven't modeled dynamic clothes yet.  the mesh looks more than regular enough to work well.

just for discussion: have you done much using the opposite effect?  using topology to create structure and invoke folds where you want them to be? and have any of you done much work with pleats?  i have a vague idea for a process for making them, but i'm still kind of stumped.

No worries, I didnt take offence or anything, I can see several places on that skirt that need a lil work :)

I have done pleats in the past, though to various levels of effect.  I hope I can explain this properly without the aid of pictures.

instead of starting with a circle I modelled each strip of skirt individualy, comeing off the waistband area.  The first step was to bring a single quad, from the wait down to the hem.  I would repeat this right around the skirt.

At this stage you have a kind of roman centurian skirt, lots of single unconnected strips, with a trangle between each one.  I then place a single tri in each gap to join everything together.

Now it depends on how you want your pleats, but a good wat of doing it is to grap the bottom line of every other quad and move then away from the original hemline, then scale them up a little to add some overlap to the underlying quads.  Then select the bottom line of the underlying quads and apply the same scale.

Lastly all you need to do is slice across the skirt to get a good mesh resolution.  Once you run the sim (with self collision on) the skirt will flatten onto itself, the overlapping polys giving a pleated effect.  The tris betweet the quad slat mean that the pleats will me more apparant the lower down the skirt goes.

Hope that helps :)

I havent done muck in the way of carving folds and then adding the mesh just yet, though I can see it would be an advantage.  One thing I have done is put a mesh through a simulation to get the folds, then take it back into my modeller to level the hem of the skirt out, so it doesn't look too wonky :)

John

hmmmmmmmm.  i think i see.  i'd have to try it out.  but if the image in my head isn't mistaken, this gives you double-sided pleat,  sort of like a gear?  not a one sided pleat, like a saw?

Quote - hmmmmmmmm.  i think i see.  i'd have to try it out.  but if the image in my head isn't mistaken, this gives you double-sided pleat,  sort of like a gear?  not a one sided pleat, like a saw?

If Im thinking right :) you could use that meshign method, then instead of movine the edges, simply move a vertex out from the skirt and overlap it with the poly below, then would give you a saw shape instead of a gear.

John

yeah, i'm kind of playing with it now and i think i see how to do it.  i still need to think about the math if i want it clean, but i think i see now.  thanks!

Quote - yeah, i'm kind of playing with it now and i think i see how to do it.  i still need to think about the math if i want it clean, but i think i see now.  thanks!

Now I think about it a bit more I think it is a question of movin the bottom verts of the tris.    If you pull one out and bring it just over the center of the tri, then pull the other one underneath it, so that the tri kinda flips over, forming the underside of the pleat.

John

If you use the common Poser type mesh you will find your garment will show creases and
distortions in a horrible way. This is seen especially where the garment may collide with either itself or where it may collide with the ground or some other prop.
Also because of the calculations involved, you will find that "chicken wire" clothing will take about twice as long to calculate than it should !!

Your mesh should not be a uniform or "chicken wire" type.
This type can be seen as nice neat squares, sometimes triangulated, but stll looking like a "wire fence".

In real life cloth has a very random pattern and that's why cloth when created for 3D must follow this randomness in order to look and perform correctly.   Delaunay mesh is a name given to this "correct" mesh and it is a simple task to convert your linear type mesh to this type.

3dsMax has a modifier that will perform this function at a click of a button.
If you don't use Max then I'm sure your application will have a similar function.

Try it and be surprised at the way your cloth looks and performs dynamically !

Cheers

"3dsMax has a modifier that will perform this function at a click of a button.
If you don't use Max then I'm sure your application will have a similar function."

I've never seen that function on any modeler.... (and I've looked at many over the years....)

Actually, although I agree that triangles will behave much better in dynamic cloth than quadrangles, I'm not quite convinced yet that randomization is so crucial. What I'd like to see is a comparison between a grid made of equilateral triangles, six around a corner, and a randomized Delaunay mesh of the same average density. Any volunteers? looks around desperately

I know mine doesn't have it. Not sure about Blender or not...haven't gotten that far. Oh, and odf, I'd show you if I could, but I use the same program you use...lmao.

Will MeshLab or PoseRay do it I wonder?

Laurie

I'm gonna look thru MeshLab and see if I can find something in there that will do it.

Laurie

negative on Poseray...
nothing I can find in Wings3D plugins.

I use C4D and I can triangulate and untriangulate. Never found anything like this. But I am a rookie with C4D

If all else fails, I could probably provide a simple, randomized mesh for testing. It would just be a flat table cloth or something like that. I've got some Java code for computing Delaunay triangulations flying around from my teaching days.

MeshLab has a Delaunay Triangulation, but I can't figure out how to use it. It turns anything I have loaded into a pile of mixed up vertices ;o). I can't do anything with it.

Laurie

Quote - MeshLab has a Delaunay Triangulation, but I can't figure out how to use it. It turns anything I have loaded into a pile of mixed up vertices ;o). I can't do anything with it.

Laurie

I have no experience with MeshLab, but had a quick peek at the documentation just now. It seems the Delaunay Triangulation filter gives you a 3d triangulation of the convex hull of your mesh. That's not what you want. The Alpha Shape function - with a suitable value for alpha - should be more useful, but only if you can figure out a way to extract just the outer surface of what it produces.

Hehehe...it's all Greek to me dear ;o). I've not been doing this nearly long enough to know what any of what you just said even means...lol.

Laurie

It can be used after clothing is made if you wish...!

If you don't use Max very often then probably read the "Help" file.

I'm off or I'll get throttled to death !!

the problem is not having max.

I can't spend $4000 on it.. I just don't have $4000 !

Yup, we have to work with what we've got...lol.

Laurie

well i can buy it but i am happy with my current tools :lol:

Personnaly i continue with quads, i have not problems with my meshs, they collide against a curvy girl not against an angular piece and are made of 7k to 20k quads.

I work in subdivision and get quads in the end, this week-end i try if triangles work for me, on Silo it's just a CTRL+SHIFT+T more

well, and i've had pretty good results with my quad meshes and those of people whose stuff i've converted.  even something with kind of bad topology like the the MFD add-ons has been OK for me.  and i'm purely in love with the Lady Littlefox meshes i've converted.  i mean, it's nice to know it can get better, but it seems like it might not be worth all the trouble it would need just to make it work properly.  you're basically eliminating the ability to add structure after the fact, and making creating structure harder.  most clothes should have some sort of structure, and if you don't give them any you get the burlap bag effect.

I have attempted to follow the "instructions" to convert a cloth mesh I have already into this "jumbled" format, but the "garment maker" is not available (grayed out) I was under the impression that is for actually making garments, not editing them.

Even so, simply triangulating a quad mesh improves the dynamics (at least in high animation scenarios)

The longer I think about this "Delaunay-cloth" idea, the less plausible does it seem to me. Most real cloth I've heard of is either woven or knitted, which means it is in fact based on a rectangular grid. I think the the 'cross-hatch' pattern that LaurieA showed earlier makes much more sense than willy-nilly randomization. I may be wrong, but I'd like to see some proof that random meshes perform better than that particular pattern, not just a straight rectangular grid.

ok, i just did some tests in Blender's cloth sim.  what i'm seeing is this:

• if you want structure, edge loops are helpful.   just putting a nice dividing line between two different topologies gave it a little more structure and made it more likely to bend there.
• tris might be better in general, but so far i'm not seeing much difference at all with a basic plane draped  on a sphere at what i'd consider medium resolution polys.  certainly not enough to be more of a concern than generally good topology.
• i used a command i hadn't noticed before called "Flip Triangle Edges" to make the topology more random and break up the grid on half of my plane.  it really changed the behavior.  not necessarily improved it, but changed it.  it made it crumple a lot more. which is great for, say, silk, but awful for say, denim.  in Poser you can control this with Sheer settings, but this will undercut it.

it's definitely not scientific, but i'd say it's still better to model in quads and control topology for structure, convert to tris if it won't kill your poly count and making running a sim a pain, and use some sort of randomizing method only if you want lots of bending and crumpling.

I think the fact that making clothing in MAX automatically uses Delaunay method is enough for me to think its the better method. that program isnt cheap!

no, but that doesn't mean it's better.  that's like saying because Max can do highly translucent SSS based skin, it must be better to do it that way.  a lot of the complaints i've read about dynamic clothes center around lack of detail.  a major aspect of lack of detail in dynamic clothing is lack of a proper structure to the clothing.   randomizing means more lack of structure.  in real life, most people don't want clothes that crumple a lot. in my experience, the conforming clothes that are more popular are either fairly structured and tightly fitted. 

dynamics simulate cloth.  a lot of the aspects i've seen complained about as "errors" are simply the simulator working like it's supposed to. when you make a dynamic outfit with no structure to it or shape it so that in real life it would gap and sag, it will sim exactly like you made it and look as poorly made as a real outfit made that way would. 

lots of cloth folds differently than it sheers.   randomizing your topology gets rid of the whole notion of direction in terms of the cloth and how it folds.  maybe the Delaunay method works better if you do some fancy simming to counteract the effects of randomization, but it seems to me like a waste of time to counteract something you don't need to do in the first place. 

that isn't to say cloth that crumples like it folds and crumples a lot can't be useful.  but i personally wouldn't use it on something like a fitted bodice or a dress shirt.

exactly the same scene, exactly the same positions and exactly the same cloth settings.

I attempted to match the numbers of poly's per plane as much as I could, but I couldnt get it exact.

In THIS example, the Delaunay method came out far superior, it looks like cloth draped over some objects, the unified quad mesh however looks terrible, it looks like I took a piece of heated plastic and let it partially melt over the objects before hardening again.

Of particular note is in the top image, one of the folds above the sphere has creased, it looks very realistic, while the bottom image is a smooth curving fold

No offense, templargfx, but I think both of those pictures look quite terrible, and not like any real cloth I've come across, except maybe spandex. From what it looks like, the stretch resistance seems way to low (Edit: or maybe it just seems plasticky because of the material settings?). Sure, that'll make the randomized mesh look a bit more wrinkly  because it has more ways to bend. But that alone doesn't make it realistic. I'm sure randomized meshes are good in certain situations, or 3dMax wouldn't offer them. But at this point, I'm still not convinced that they are right for most types of woven or knitted cloth.

And people, you need to test things in Poser if that's what you'll be working with in practice. Cloth simulation is ridiculously complicated. Your results will depend significantly on the implementation you're using. If Delaunay works best in 3dMax, that does not mean it will in Poser, which for all I know could be optimized to deal with quadrangles exclusively. Besides, if someone can dish out several thousands of dollars for a high-end application, they might in fact use specialized add-ons for things like cloth simulations if and when they really need them to work, and never touch the one in the base program at all. So, the "it must be good because it's expensive" argument is not necessarily something I'll buy into.

What I'm saying is, we need to do the research, or we'll just keep guessing. We need to do it in Poser, and we need to use parameters that make sense for the kind of cloth we are interested in.

I'll see if I can whip up some test meshes with a range of regular and randomized topologies over the weekend, so you guys can do some serious testing if you want. I'll have a play in the cloth room myself, but I'm not well-versed in the art of setting up a simulation and picking the right combination of parameters.