Sim Cloth Testing

[Paul Forwood]

The project that I am working on requires alot of dynamic interaction between layers of cloth and hair so I have been testing these features recently. I decided to start a thread to show what I find and hopefully get some answers on the way. I don't know if there are any differences between Sim Cloth in A:M 13 and A:M 14 but I am testing both.

 

This is a 2 second movie, from A:M 14 Beta 4, which shows that Sim Cloth on Sim Cloth is possible:

 

The missing patches on frame 0 is an A:M 14 bug, I think. It always has a render problem on the first frame of the QT movies that I have rendered from Beta4.

 

I have noticed that when I render to Sorenson 3 format from A:M 14 that along with the QT movie I get a "sinfo" file. Does anyone know what these files are and whether they can be turned off and deleted? I am assuming that A:M 14 is creating these sinfo files for it's own purposes as I have never seen them before the Beta 4 release.

 

I am still not at all clear about the practical purpose of most of the Sim Cloth properties. For instance what is the result of increasing, or decreasing, the mass property of cloth? The only thing that I can determine so far is that if I crank it up much above 5 I get the "unsolvable" error message.

[KenH]

From experimenting, increasing the mass seems to make the cloth react more violently ie air resistance has less of an effect on it. Sometimes I think you have to close/reopen to see changes on some of the values.

[Paul Forwood]

From experimenting, increasing the mass seems to make the cloth react more violently ie air resistance has less of an effect on it.

 

So does mass = cloth weight? If mass is a control for air resistance then what is "Air drag" for?

 

Sometimes I think you have to close/reopen to see changes on some of the values.

 

Yes. I've noticed that too. So to be sure that you're seeing what you're working on you have to save and then reopen the project every time you adjust any of the properties. Hopefully finding some good cloth settings for various types of cloth will enable people to avoid all this tweaking, saving reloading, simulating over and over and over again.

[noober]

Sometimes I think you have to close/reopen to see changes on some of the values.

Have you tried hitting the spacebar? It updates the window your working in. I know with hair it will calculate new changes.

[Paul Forwood]

Have you tried hitting the spacebar?

Yes but it doesn't work for everything.

 

The following example is another test I did in A:M 14 beta 4:

 

 

I made a few adjustments after this and rerendered. The cape developed a glitch and wouldn't drop down from a raised position, even though the mannequin was stationary.

 

 

I saved the project and reopened it and noticed that the mannequin was inside out so I flipped the normals, adjusted a few CPs, made some adjustments to the hair, made some changes to the choreography, ran the simulation and it looked good. So I saved the project, got some exceptions and finally a crash. Rebooted A:M, loaded the project and the model was completely trashed. (hmmm. someone has removed text from this post!)

 

Going back to A:M 13 for these cloth tests until the next Beta arrives.

[MMZ_TimeLord]

Paul,

 

Looks like the cape got stuck in the hair on test A09.

 

Good experiments though... great job!

[robcat2075]

I haven't done much with hair, but I've sometimes wondered if something could be gained by scaling the time after a simulation.

 

Like maybe the hyperactivity of the cloth could be reduced by animating my character for five seconds, stretching that out to 10 seconds, setting the cloth to half its mass, simulating it, then scaling everything back to 5 seconds.

 

Not that that is the right recipe, but maybe time scaling might get an effect not possible otherwise.

[Paul Forwood]

Thanks, Jody!

 

Robert, I know what you are saying but I would like to avoid looking for awkward workarounds if possible. I would just like to see Sim Cloth nailed down so that we can understand better how to use it and to predict it's behavior when making adjustments to it's properties. A TechTalk that goes through each of the properties would be very useful but I don't expect that is a priority for Hash Inc at the moment.

 

Here is a test that I tried in A:M13s:

 

 

 

I think that it is pretty amazing how Sim Cloth handled this test! It uses two different cloth settings. The cape uses the default Sim Cloth settings while the cloth squares that are dropped onto the mannequin all use the following settings:

 

Mass = 1

Self Collide = On

Friction = 1000

Stretch/Shear

Stretch Stiffness = 1

Shear Stiffness = 1

SS Damping = 0.1

Bend Angle

Bend Angle Stiffness = 0.1

Bend Angle Damping = 1

Air Drag = .001

 

The mannequin and the floor are deflectors with the subdivisions set to 16.

 

If I can keep the flexibility of this cloth while avoiding all the stretching at the top I will be getting close to something useful for my current needs though there are still plenty of other tests to be done to see how folded seams will hold up and tests to do with textures and hair. Also I would like to build a small selection of cloth types and test them with things like Forces.

 

Does anyone have a good explanation of how damping works? Okay, I understand, from reading Bob Croucher's Sim Cloth thread, that it controls the amount of resistance that the cloth has to deforming when undergoing changes in speed and direction. (At least that is how I interpreted it). But does it have complete control over that feature of cloth or does it share control of deformation, due to changes in velocity, with other Sim Cloth properties? More testing needed.

 

------------------

[MMZ_TimeLord]

Wow... five layers and only a little bit of break through... nicely done!

 

Some tweaking on the sub step calculation may help lessen or eliminate that.

 

As to the cloth stretching, probably increasing the stretch stiffness, but I'm not sure the effect that will have on the rest of the simulation.

 

Keep going Paul, I look forward to more of your experiments.

[Dhar]

This looks incredible, Paul. Nice work.

 

I can't seem to be able to choose a Custom Group. I have a model which is basically a circular mesh with the innermost ring group of CP's named "Attached", but when I go to the Custom Groups in the Attributes, there is nothing in the drop down list

 

Any ideas?

[Paul Forwood]

The way that I have been applying cloth is by dropping the cloth and deflector materials onto groups while in modelling mode. I leave the materials' properties in their default state and only adjust the properties of the materials within the groups. I haven't had any problems with attributes vanishing.

[T-Dogg]

This looks incredible, Paul. Nice work.

 

I can't seem to be able to choose a Custom Group. I have a model which is basically a circular mesh with the innermost ring group of CP's named "Attached", but when I go to the Custom Groups in the Attributes, there is nothing in the drop down list

 

Any ideas?

 

Dhar,

 

It sounds like you might be clicking on custom groups for the cloth material itself. The custom groups for the cloth in MATERIALS is always empty. If you go to where your cloth is applied to your object, you should be able to click on custom groups and see a list of all the groups. Your attached group should be visible there.

 

Hope this helps.

[Paul Forwood]

Well I have been continuing my Sim Cloth tests but have not really progressed much. So much trial and error. So little to show for it.

Dropping a cloth square onto another object is probably one of the easier things to do with Sim Cloth. Once you start using it with more complex objects it produces lots of "unsolvable" errors. Also getting a cloth material that will conform to the underlying shape and adding motion is just escaping me.

 

I have a better grasp on "Mass" now though. It seems to simply set the relationship between colliding CPs, indicating which one has the greater force.

 

Here is another A:M13s test:

 

 

 

I'm going to go back to very simple objects and try various mesh densities next.

 

--------------------------------------------------------

[agep]

Hahaha! your last test was brilliant... nice texture

[oakchas]

What FUN! that last one looks as believable as can be!

[Paul Forwood]

Some tweaking on the sub step calculation may help lessen or eliminate that.

 

Sorry, Jody. I missed this earlier but having seen it now I 'm not sure what you mean. What is the sub step calculation?

 

I may be using Sim Cloth incorrectly. I am just creating groups on the model. I drop my cloth and deflectors on these and do any property tweaking in the model window. I then jump back to the choreography, run the cloth simulation and if all goes well I render out the test to a QT movie. Is this the correct way to use Sim Cloth?

 

I have watched Bob Croucher's TechTalk on TWO cloth, which incidentally was made exactly one year ago today and which was very helpful, but it is refering to an older version of Sim Cloth. The current version does not have the same properties as the version in the video. I wish the programmers would document the features that they implement. It's all guess work otherwise. I spend more time trying to find out the things that should have been documented in the first place than doing anything creative.

 

Some questions:

 

1) Bob Croucher mentions something in the TechTalk about keeping cloth at least half a centimetre away from deflectors. The current version dosn't have a configurable collision radius so what is the mimimum safe distance between cloth and deflectors?

 

2) What is it that makes cloth suddenly jump out and how can that effect be dampened or removed?

( See attached: )

 

3) Do normals affect the way cloth surfaces react with one another?

 

 

 

Hahaha! your last test was brilliant... nice texture

 

Thanks, Stian.

[robcat2075]

Is the character made to scale? Is he normal adult size?

[Paul Forwood]

What FUN! that last one looks as believable as can be!

Thanks, Charlie!

 

Is the character made to scale? Is he normal adult size?

The toony character is a slightly modified version of Zaryin's Squetchy Sam model, rigged by David Simmons. He stands 157cm tall, (approx. 5'2").

The robed and caped character is quite tall at 203 cm, (6'8"). This one isn't rigged yet. So yes they both conform to "normal" adult sizes. Neither have been scaled in the choreography.

[Dhar]

Paul, adding "Attached" to the Custom Group seem to work in v13s but not 14Beta4. Do you have the same issues?

[zandoriastudios]

Paul, I did some experimenting myself tonight....

1. Collision radius seems to be controlled by the collision tolerance.

2. denser geometry drapes better...

3. Do not use HOOKS--they don't seem to simulate (notice the piece sticking up in the back). I reworked the mesh a couple of times to figure out... 5points are ok, so are 3 points, but a hook throws the simulation off.

 

I ran this simulation after playing with the settings. everything is default except the collision tolerance. I was pleased, though it is a little jiggly at one point ( I think increasing the damping would fix...)

shirttest.mov

[Paul Forwood]

Paul, adding "Attached" to the Custom Group seem to work in v13s but not 14Beta4. Do you have the same issues?

 

Dhar, I had to do some searching through the forums to find a definition for "custom groups". This doesn't appear to be used in the versions of A:M that I have been testing Sim Cloth with, (A:M 13 and 14 beta4). I finally found this excellent thread which helped me understand where much of my confusion is coming from: A:M13 Sim Cloth

 

I had not tried Sim Cloth until now and so many of the terms that people were using, such as "attachment, custom groups" and various properties were foreign to me. They seem to have been removed in the current version of Sim Cloth. In A:M 13 and 14 Sim Cloth is applied just like any other material. There are no custom groups or attachments that I can see. I like this straight forward approach to applying cloth but it seems that some control has been lost in the transition. Reading the thread above it seems that the "attachment" group was important in adding stability to cloth. I would go back to A:M 12 but I started this project in A:M 13 so there is no way back for my models.

 

Will, what version of A:M did you do that test in? There isn't a "collision tolerance" property in A:M 13 or 14 that I am aware of. If you know better please show me where it is hiding.

Yes, I confirm what you say regarding density, 3 point patches, 5 point patches and hooks. I expected problems with 3 and 5 point patches but I haven't noticed any yet.

[zandoriastudios]

V13. the collision tolerance is in the Choreography properties

[Paul Forwood]

Ah! (light goes on)

Thanks, Will!

[Rodney]

What is it that makes cloth suddenly jump out and how can that effect be dampened or removed?

( See attached:

 

Paul,

It would be interesting to see what areas are defined as your deflector groups under the cloth.

In areas such as under arm pits, groin vault etc. you most likely will not need any.

 

There are some examples here in the forum where the underlying geometry is entirely removed.

This then dampens the cloth automatically as there is nothing to deflect in the areas of built up geometry.

Was it Xtas that did those?

 

The principle goes back to not modeling what won't be seen.

If there is no geometry there there can be no pass throughs and nothing to drive strange deflection.

Just make sure there is adequate deflection in those areas that need it.

 

Note: I'm not sure if just hiding geometry will help but that would be easy to test.

[Paul Forwood]

Will, thanks again for pointing me to those properties. They make ALL the difference!

 

I expected problems with 3 and 5 point patches but I haven't noticed any yet.

 

I discovered that 5 point patches are indeed a problem!

Where I was having problems with CPs jumping out, sometimes enormous distances, I noticed the cloth patches where always sitting over a 5 point deflecting patch. By removing a few CPs from the deflector group and the cloth group I was able to remove that massive distortion completely. So, as I originally suspected, 5 point patches should be avoided on the deflectors, just like hair emitters . (At least that is what I am seeing at the moment).

 

The SimCloth properties in the choreography are so important in getting the right control over the cloth and that is what I am going to start playing with next. Collision radius was set too high by default for close fitting cloth so I turned that down to .5 and so far so good.

 

Here is the last test:

 

 

 

Lots to tweak and test. He needs a new gown too. This one was designed to be a static chest without cloth so the splines are not running the right way.

I also want to put the Squetch Rig into this model.

 

I must say I am so encouraged by seeing the cloth calm down, and by finding that 5 point trouble maker, and by having those illusive Sim Cloth properties pointed out to me that I feel all positive about SimCloth again.

 

Thanks, everyone, for the help!

---------

[Vance]

It's really come a long way! Congratulations.

 

The cloth still looks just a little bit too springy to me. It has a little bit more bounce than real cloth does unless it contains rubber. Other than that, it is really getting there!

 

-Vance

[Paul Forwood]

There are some examples here in the forum where the underlying geometry is entirely removed

 

Rodney, this model was built in that manner originally. The only cloth that I was going to use was going to be on the cape and the bottom of the gown, from just below the hips. I was going to use dynamic bones for the sleeves and the arms only went just past the elbow. I will probably still do a version like that but after seeing Will's test and the examples that I have seen in the thread that I posted a link to, above, I was determined to at least try draping a full body to see how Sim Cloth would handle it. The underlying geometry can be deleted at any time that I decide I don't need it. This is a very low patch model and I want to keep it that way as far as possible. I know that SimCloth will perform better with denser meshes so there will be some splines added when absolutely neccessary.

 

It's really come a long way! Congratulations.

 

Thanks, Vance.

 

The cloth still looks just a little bit too springy to me.

Agreed, but I have hope that a half decent result is attainable now.

[Paul Forwood]

Just another bad cloth day:

 

(blooper)

 

I just threw a TSM2 rig in there and haven't done any of the intermediate bones, or any proper animating, so it looks strange. I just needed to start testing Simcloth with some more typical motion.

 

The distorting CPs have got me puzzled at the moment. I thought that I had eliminated that by removing 5 point patches from the cloth groups but it is back with a vengeance.

 

I'm going back to just using cloth for the bottom part of the gown and the cape. The sleeves and hood will have dynamic bones and the rest will not have dynamics. This is probably the best solution for a faster and more robust workflow.

 

I'm looking forward to Rich (Entity) Jackson's forthcoming Simcloth documentation.

[Paul Forwood]

Something strange has occurred in A:M13s!

 

If you look at the attached image you can see that the patches have corrupted somehow. The change occurred while I was checking the normals of the Squetchy Sam model. I found that there was on patch on his back which was facing the wrong way so I flipped it. All the normals are facing out now but look at the state of the model in a choreography! The model looks okay in the model window but when I look at it in a choreography it is corrupted. Rendering doesn't improve the image either.

 

 

 

I have reinstalled A:M13s but that doesn't make any difference.

 

Has anyone else seen this? Are there any known fixes?

------------------------------------------------------------------

EDIT:

It seems that something corrupted the Rig when the normal was flipped. Is this likely?

I have tried creating a new action but the FACE rig produces bad results. Not sure if this is all related to Simcloth yet.

 

 

 

Oh, by the way, the reason that my eye was drawn to the flipped normal was because a CP on the SimCloth was being massively displaced directly above it. It seems that normals DO have an affect on cloth.

[itsjustme]

What it looks like to me is that the cloth was applied in another action and then when it was put in the Choreography with the character, the CP numbers that were applied to the cloth were the same as part of the model. My guess is that you would need to apply the cloth in the same Choreography or Action as all of the objects that are going to be in there so that there isn't any cross-assigning.

 

Hope that helps, Paul.

[Paul Forwood]

Hmmm. That gives me something new to think about.

 

The choreography has just two objects in it, (the ground object and the character). The poncho, the hat and the cigar have been added to the Squetchy Sam model's geometry and 8 CPs around the neck of the poncho have been assigned to the chest geometry bone. The hat has it's own bone which is a child of the head_back geometry bone, (or similar name), and the cigar's bone is a child of one of the lip bones. Perhaps when I did this parenting it messed something up?

 

The cloth materials were applied to groups in the model window.

The additional geometry was added to the model after the action was created which might have caused problems as it was copied and pasted with cloth material already assigned.

 

I guess that I shouldn't be surprised that something broke. Oh well. It was just another experiment.

 

Thanks, David.

[mtpeak2]

That does sound like it would cause problems. Try removing the cloth material from the groups and reapplying.

[zandoriastudios]

another approach might be to make the poncho a separate model. then create an action to dress the character. In the example that I posted, the shirt was separate from the Thom model. This approach would make it easier to have different wardrobes for a character.

[goodguy20k]

Just curious here, don't want to hijack the thread but...

 

With your testing, Paul, what do you think of "soft bodied" models with the current simcloth setup? Basically, creating a "hard skeleton" and then throw fabric over it. I've thought about testing this, but... Never had the time. I'm curious what you think about it.

 

Very cool tests, and I'm watching eagerly!

[Paul Forwood]

Try removing the cloth material from the groups and reapplying.

Thanks, Mark. I tried that but it didn't work. I still have a clean version of the model so I will continue trying various cloth settings and keep looking for the things that cause the problems as well as the things that work well.

 

another approach might be to make the poncho a separate model. then create an action to dress the character.

Yes, I am going to try that one too. Thanks, Will.

 

what do you think of "soft bodied" models with the current simcloth setup?

Still too early to say, Daniel. I can see lots of potential for certain things but I am finding it difficult to get full predictability from the little that I have learnt while testing cloth.

 

I have just been trying various simcloth settings on a simple sack model walk-cycle:

 

 

 

 

The top two rows are all empty sacks while the bottom row uses one internal deflector for the first sack and the last two use two internal deflectors. I should mention that the middle row had Air Drag set to 0.1 also.

No major breakthroughs to report yet.

[Paul Forwood]

Here is the result of my most recent test of cloth:

 

 

 

The animation is just some rough movements to shake the cloth and hair about and put it through some of the challenges that would be required in a production. I think I will build a little obstacle course to put the characters through so that the dynamics can be better tested.

 

The choreography consists of three actions which follow each other with no overlap. (They should actually overlap by one frame). Action 1 is just a rough twist and bend to test the reaction of cloth when it touches the ground plane and to test for penetration of knees through the cloth. Action 3 is a rough walk cycle. Action 3 is a new action which acts as a transition between the other two actions. The last frame of action 1 is copied to the first frame of this new action and the first frame of action 3 is copied to the last frame of this new action. The character is constrained to a path at frame 0 but the enforcement and ease values are not active until the start of the walk cycle. It worked well but I can see that I missed matching the keys for a few bones.

 

I didn't know that action to action copying was working at all until now. This is major! (Note: It didn't work flawlessly. I had to keep trying, doing a copy on the hip controller with filters set to "Key Model" and selecting that bone in the new action before pasting gave me some mixed results but maybe I should have been copy/pasting a root bone. I did end up copying some keys individually. Not sure but I guess it depends on the rig. The rig in this model is TSM2). I also notice that my model has a problem in her right shoulder.

 

Oh, by the way, I should mention that this was done in A:M 14 Beta 5.

[Mohammad]

I'm working with cloth in my animation too, I'm using an older version not sure if it will be as good as yours. The A08 comp looked really food though!

[Dhar]

Wow Paul! That looks great. Can I see the wireframe of the dress? And what are your cloth settings?

[Paul Forwood]

Can I see the wireframe of the dress? And what are your cloth settings?

 

Here you go:

 

 

The cloth is still too stiff at these settings but it is the loosest that I have been able to achieve so far without having a leg penetrate the cloth.

[robcat2075]

that's quite promising looking!

 

The cloth is still too stiff at these settings but it is the loosest that I have been able to achieve so far without having a leg penetrate the cloth.

 

is it possible to make the legs transparent, so the pass throughs wouldn't show?

[Paul Forwood]

is it possible to make the legs transparent, so the pass throughs wouldn't show?

 

Yes. That is possible, as a last measure, but I would prefer to see what I am doing at the moment. It is preferable to keep the legs from penetrating or the cloth can get snagged inside them which can end up pulling more and more CPs in, as you can see here:

 

 

Making troublesome parts transparent could be a very useful workaround though, so long as that part would never be seen anyway.

[steve392]

The dress works good and the hair moves well ,if you could get the legs not to be seen passing through it would be marvelous ,nice rig you got there to get that movement Paul

[Ganthofer]

I experimented a while back with Cloth and you might want to try making some denser meshes in the problem areas. I found that I could use the denser mesh during the simulation and switch to the normal mesh during rendering. If you don't want to have 2 versions of your model, you could create some mesh shields and constrain them to the legs and remove or hide them when you render.

 

Just some ideas that you might find useful.

 

Glenn

[Paul Forwood]

nice rig you got there

Thanks, Steve. It's The Setup Machine 2. I badly need to add some shoulder controls to this rig though.

 

I found that I could use the denser mesh during the simulation and switch to the normal mesh during rendering. If you don't want to have 2 versions of your model, you could create some mesh shields and constrain them to the legs

 

Thanks, Glenn. Yes, denser meshes are in the next revision but I am determined to keep these models as low-patch as possible and find workable cloth properties for each level. It will hopefully all get better as the density of the surface meshes increase.

[robcat2075]

When you get cloth all figured out, Paul, I'd sure like to know how to do the sort of subtle, non-lively cloth we see on clothes such as these:

 

 

Accroding to he promo material those suits are simulated cloth and not explicitly modeld that way. I realize lots of R&D has gone into Pixar's cloth simulator, but maybe A:M cloth just needs to find the right settings and tricks to get the same result? I'm sure there's lots of tricks and workarounds in a Pixar movie.

 

Anyway, keep at it Paul, and please do a tut when you're done that explains what you've found out.

[Shelton]

Paul

 

Excellent study here. I would be interested in a tutorial as well

 

Steve

[Paul Forwood]

please do a tut when you're done that explains what you've found out.

Sure. I'll do that when I have some good examples that have been put to the test.

I tried the settings that Holmes recently posted in his tutorial and they worked though the cloth was much stiffer than the example that he showed.

 

I have got this slightly more flexible setup without any penetrating sufaces. I will post the settings later.

 

 

 

This simcloth would just about do for what I have in mind but now I need to get a cape to rest on top of the skirt and under the hair.

[Ganthofer]

Just my 2 cents here.

 

I would run it once with the skirt (no cape or hair), then make the skirt a deflector and run it with the cape (no hair) , the run it with the Hair ( with the skirt and the cape already keyed ). I think this would be best, although you obviously have to run 2 cloth sims, but since there really wouldn't be much interaction between the skirt and the cape it should work and the cape will drape itself over the skirt. Of course, you would need to get the skirt animated the way you want it before running the sim on the cape.

 

Oh, and by the way, very nice cloth skirt movement. The model is moving a little jerky though.

 

Glenn

[Paul Forwood]

I would run it once with the skirt (no cape or hair), then make the skirt a deflector and run it with the cape (no hair) , the run it with the Hair ( with the skirt and the cape already keyed ).

Hmmm. That is something that I haven't tried but I'm still hoping that I can avoid that extra step. Good suggestion, Glenn. I will keep it in mind if everything goes pear shaped.

[Paul Forwood]

Here are some more examples and the settings used:

 

 

 

 

 

These are quite subtle variations. Next step will be to increase the density of the skirt which will probably lead to increasing the density of the legs too.