nCloth starter

[mikael persson]

Hello list,

I am about to start on a project where I have to recreate fashion shots of realistic clothing. As I haven't used nCloth much before I was wondering if anyone of you out there might have some good hints before I dive too deep into it.

I know that the clothing is supposed to be built as close to real clothing construction as possible to get a realistic look.

-Do you create NURBS surfaces out of boundary curves and attach the resulting surfaces together? Or is it better to go straight to polys?

-What is the best method for creating bunched up clothing, like sleeves at the elbow?

-When you want alternating thickness, like thicker bits close to the hands. Do you create two meshes and two nucleus objects that you constrain together?

-For the thicker parts is it best to create the actual thick geometry with polygons at both sides, or just set the thickness for the solver to a lot more than the ordinary fabric?

Any pointers are greatly appreciated.

Mikael

[Duncan Brinsmead]

For nCloth I would suggest to model a uniform quad mesh.  Keep it as absolutely simple as possible and sooth the mesh downstream of the simulation. That, or use it as a wrap deformer on more complex geometry. Thickness is a paintable cloth attribute. In terms of the render you can extrude downstream of the simulation to thicken the mesh visually. (you don’t want to simulate on the extruded mesh)

 

Modeling can be tricky if you want all the uv seams to exactly match the way a totally flat printed fabric would assume a shape. The panel modeling workflow of the old classic cloth was nice in that the modeling preserved these properties, but it was cumbersome to use.  I suppose in nCloth one could start with flat planes, create edge constraints(setting the constraint rest length to zero) between them and simulate, colliding with your chacter. Then do poly combine and merge vert on the resulting meshes(sort of modeling using nCloth). This sounds like a world of pain to me though. Standard modeling with eyeballing the look should be easier for most garments.

 

You may wish to get bunching effects through simulation, rather than trying to model them. In some cases this is simply animating your character from a rest  pose to a state where bunched, then setting the initial state.  In others you may wish to run simulations with collisions and animated constraints to pull clothing into shapes and folds, then copy the cloth output mesh to create your base mesh(and delete all the old temporary cloth simulation). Certain types of bunching shapes are much easier to get through simulation than modeling.

 

Duncan

[ma...@tokeru.com]

There's a tutorial somewhere explaining how to use elements of the old
cloth workflow with ncloth, including a few helper scripts to make it all
go faster. I've lost the link though... :( Anyone have it?

Written in english, but the author was japanese or korean from memory. Was
making superhero outfits, tight fitting lycra wonderwoman style things
(ooo-er).

You'd think that'd be an easy google search, but I came up empty handed.

-matt

<snip>

> Modeling can be tricky if you want all the uv seams to exactly match the
> way a totally flat printed fabric would assume a shape. The panel modeling
> workflow of the old classic cloth was nice in that the modeling preserved
> these properties, but it was cumbersome to use. I suppose in nCloth one

<snip>

[Duncan Brinsmead]

One way to do tight stuff is to model the shape a bit loose then lower the restLengthScale on the cloth while simulating to tighten. When doing this it sometimes help to lower the stretch resistance to keep it from fighting the collisions too hard. For a more controlled simulation it is also generally better to animate the rest scale lowering rather than simply setting it low.

If you are just getting started with nCloth the following tutorial should be useful. It covers some basic setup issues that people commonly hit.

http://area.autodesk.com/blogs/duncan/basic_cloth_on_character_setup

Duncan

[pha...@cox.net]

I think this is script you're talking about Matt.
http://sites.google.com/site/cgriders/jc/clothes/

It's pretty cool. The one issue I noticed is that the number of verts at the seems of panels don't match up. Creating the geo by way of the surfaces>planar tool is the problem I believe.

-Paul

[ma...@tokeru.com]

Ding! We have a winner!

Thanks Paul.

-matt

[Duncan Brinsmead]

Ha, I didn't know someone had put together a pattern modeling tool for nCloth. Stills seems like a complex way of working( unless you need to match a pattern ), but I know it is important for some and it is cool that this is freely available.

Duncan

-----Original Message-----
From: maya...@googlegroups.com [mailto:maya...@googlegroups.com] On Behalf Of ma...@tokeru.com
Sent: Tuesday, February 09, 2010 9:44 PM
To: maya...@googlegroups.com
Subject: RE: [maya_he3d] nCloth starter

[mikael persson]

Thanks all for the helpful pointers.

After two days of hard simulation I've learnt a lot more, but still have a few issues.

I found painting the different variables very helpful, especially the wrinkle one.

But my main problem is still the bunching of clothing. When I pull up the sleeve of the garment to attach it just under the elbow it creates too much intersecting surfaces.

I was hoping that the cloth would bulge over and create nice rings. Is this just a matter of resolution of the mesh? Or very high values for the collision?

At the moment I used the simulation as a starting point then went in a hand modeled to get a good look.

Another thing is finding the balance between speed and accuracy for the simulation. Sadly I've found that I need to crank up the accuracy quite a bit to get good results, but then there is a wait of a couple of minutes to test each value change.

Do you guys have any hints in that area?

Mikael

[Duncan Brinsmead]

The first thing to try in increasing substeps (the default is quite low for good collision quality). Also be aware that the cloth  and passive meshes have thickness for collision and sometimes pulling things closer than the thickness can cause fighting between a constraint and collisions. Make sure than any constraints you create do not overly bind the cloth. If you want fine wrinkle details then the mesh resolution needs to be higher( as well the substeps and stretch resistance). Check that there is no self intersection at the start frame.

 

Duncan