Cineversity

Hi there

I’m struggling with a cloth sim and cannot find out what is going wrong. I’ve got my character rig and set up with mocap data from mixamo- The dress is a cloth sim attached to the model with a belt tag. On the model itself I’ve got a cloth collider tag. Within this tag I have also determined which polygons to ignore for the collision.

When I run the sim, the legs of the model interact the way they should with the cloth, but the bum of the original mesh somehow holds the cloth back; so I’m confused because the front legs work fine but the bum does something else…if that makes any sense.

Any ideas?

Cheers

Steve

Here’s the link to the file:

https://www.dropbox.com/s/a821sbizujrd856/annie_cloth_dress.c4d?dl=0

Cheers

Steve

Hi Steve,

Thanks for the file.

While opening the scene, I saw that the rock-disc was kind of way behind her. The animation starts later. That is the part I will focus on.

The movement of the Belt-sphere and the Spine joint will not match the area where the belt is.

A copy of the rock disc and the belt skirt (on the pale green layer) was set up newly. The Belt base sphere was shaped more to fit the body.
Please have a look at the Expert Tab in the Cloth Tag. Perhaps adjust the values there. The Subsampling is undoubtedly the main parameter here.
The Hover values are set to 0%.

I relaxed the rock disc only a little bit, as the geometry as given will intersect very quickly otherwise.

The animation starts later, I have excluded the earliest parts.

I have attached a camera on the Belt-sphere, to have a better observation of what happens.

I hope that comes close.

Scene file
https://www.amazon.com/clouddrive/share/ioFSQNZqt2XldYgXEL4scK9KorlvGxLVFcMHHNhsow7

All the best

Hi Dr Sassi!

By the way: It’s a total coincidence that the walk cycle is called ‘sassi’!

Thanks for the file. I already had a quick look into it, will have more time this evening.

So you suggest that the main problem was the jolt at the beginning of the animation? So it would always be best, with simulations, to avoid that jump before an animation sets in?

Thanks a lot! Spent quite a lot of time already going through possible mistakes!

Cheers

Steve

HI Steve,

You certainly had good intention mentioning it. Between sassy and my name is a big difference; I’m happy when people keep the difference.

When you ask about the main problem then I have the feeling I have to go deeper into it, it is not a single thing. I’m not sure if I got all, as at one point, I just took copies and started freshly over.

I can’t really tell what the problem was based on a single item in the scene. When I opened the scene file here and played it from the start, there was a massive jump from the skirt towards the character. Which introduced a lot of dynamics. Of course, that can be on purpose, but that is not something that leads to an initially stable setup.

The Belt was made from a sphere, not adjusted to the shape of the needed Belt. Furthermore, the joint used moves, not the parts obviously where the Belt was. With the joint to the sphere to belt to skirt, that is a different information flow, which might work similarly.  I’m not clear why the sphere was used, to have additional animation options? The Belt’s parameter of Hover has undoubtedly an influence of the ‘swimming” of the skirt. Dial it down if the sphere is shaped along the belly area.

The geometry is pretty rough for the skirt, and I thought to have some colliders during the relaxing, to support its folding, but that is already art directing, and I got no sign that the shape you go was not wanted. Kai had made a video about an idea to get a better cloth.
https://www.cineversity.com/vidplaytut/better_cloth_deformations
Again, this is art directing, and your call, of course.

The Expert Tab needs your attention, the default values are often sufficient, but mostly only close, and of course, sometimes not at all. One can slow down the system (at least mine) to nearly halt with too much calculation, based on those settings. After all, any “simulation” in 3D is a try to get away with a minimal fraction of real physical simulations. It is pretty much a simplification, nothing more.

A clean set up, like shaping the sphere, or modeling a proper belt if the original surface shouldn’t be used, is a good start. You excluded geometry that is not needed, which certainly helps a lot to cut down the calculation or have a better one with it, given the same amount of time.

I’m sure you will get there, and going through those things a few times will give you the certainty to be faster and quicker each time.

My best wishes.

Hi Dr Sassi!

I found out what the problem was!!

Hidden in a Null that was set to ‘Visible in Editor/Render’ as Off I still had my ‘backup’ of Annie with another Collidertag on it. So that was messing up the sim! Now everything works just fine!

I did however remove the sphere I used to attach the belt-tag and moved that to the original mesh of Annie. I’ll still do a bit of cleaning up now, but I’m really starting to get the feel for how this cloth sim stuff works; thanks to you aswell!

Cheers

Steve

Hi Steve,

Thanks for the feedback. You have obviously sent a file that is different than what you are referring to. I downloaded the file of yours again to check if I had deleted or disabled any other Tag. This was my first check, and it is the same with the second download:
https://www.amazon.com/clouddrive/share/DfRKnr194EdJOBENROeJOiB7cAOjjXpll9goQb2WAWx

I disabled the Dynamics tag, but more out of routine (see file above, post #2)

Have a look here, there is no influence from other collider Tags.
https://www.amazon.com/clouddrive/share/oitarUrUaCk3cL6Guz9rYXfxeILK96NSCOzzSwRFQhX

I know that sounds like “I need to be right here,” and I hope I’m above such low behavior, and I have not to endure any “projections” now. My intention here is to not let you believe that you can’t mix those it with other dynamics. Hence my effort to set up the file above.

If you had another scene file that I had no access to and there was an additional Cloth Collider Tag, then yes, of course, that will limit the fun, if that was unintentional. But the scene I had, it was not, and I can’t leave this as valid information here, it might limit other readers fun.
I hope my reply makes sense.

Since we are in a forum, as mentioned: One step that I do often is, that I copy the absolutely needed files into a new scene, and see what happens there. (I usually ask for a reduced version of the problem files, as such a reduction often leads to a problem elimination work by itself.) Perhaps worth a try if you hit a problem at all again. Fingers crossed you don’t need to.

Thanks for your reply, I hope it was worth it.

Cheers

Cheers.

Hi there

You are actually right! I did delete quite a few things from the original file before I uploaded the one you got. Along with the mentioned null aswell.

Anyhow, the problem I was having did actually come from that additional collidertag in the hidden null. The issue with the cloth jumping at the beginning wasn’t actually part of my problem (I also set the animation about 50 frames back to let the cloth fall first before the actual animation should set in). I changed that now, because as you mentioned it did add additional dynamics to the scene which made things a little hectic.

Thanks for the help!

Cheers

Steve

Hi Steve,

Your reply makes me happy, as I can see that we are both predominantly interested in solving things. Thanks for the extra time you have invested, very much appreciated.

My best wishes for your animation-project.