Forum
Welcome, Guest
Username: Password: Remember me
  • Page:
  • 1

TOPIC:

Stitch constraints and maintaining drape from Marvelous Designer 2 months 11 hours ago #794

  • slhouette
  • slhouette's Avatar Topic Author
  • Offline
  • New Member
  • New Member
  • Posts: 6
  • Thank you received: 0
Hello again. I'm a bit confused how to tackle this problem with my mesh - I'm pretty sure it's an issue of the stitch constraints not holding. So I have two skirt layers that are both draped like so, coming from Marvelous Designer: 


The resulting sim flattens out all the pattern pieces and there's a really unfortunate effect of the back panel falling completely flat.   
 

I've had this same problem before in a Houdini project and corrected it by increasing the stiffness of my welding constraints - in Carbon this would be my stitches. However, my main issue is that I'm not getting any results back from changing my parameters. I've really raised my stiffness in bend/stretch + stiffness of my stitch constraints, and really bumped up my subdivisions and iterations to compensate, but I'm not seeing any change at all. I imagine this might have something to do with my mesh density, which is approx 29000 for both meshes. My subdivisions and iterations are 100 and 120 respectively - To start, is this an adequate amount for the number of verts? 

Thanks! 
Attachments:

Please Log in or Create an account to join the conversation.

Stitch constraints and maintaining drape from Marvelous Designer 2 months 1 hour ago #795

  • Sebastian
  • Sebastian's Avatar
  • Offline
  • Administrator
  • Administrator
  • Posts: 112
  • Thank you received: 22
Hi slhouette.

May I ask what you are using the Stitch constraint for specifically?
Does each dress layer come as a separate mesh, is it all in one geometry, are you using the Garment Processor workflow, etc.

Also, are you applying flat reference meshes to the Carbon Cloth nodes?
If you use flat reference meshes, the simulation will try to conform to this reference, i.e. will flatten any angles/meshes.
In that case, increasing Bend Stiffness on the Cloth and Iterations/Subdivisions on the Simulation will have the opposite effect of what you are trying to achieve, i.e. the Cloth will get closer and closer to the flat state, i.e. convergence gets better and better. (see below more for information on this).
If that is the case, maybe try to not use flat Cloth reference meshes and see if that gives you the look you want, i.e. you can just remove the reference meshes from the Cloths and only have Start Meshes. In this case, it would keep the look / shape from the Marvelous meshes.

I imagine this might have something to do with my mesh density, which is approx 29000 for both meshes.

Is this 29k for each mesh, or 29k combined? Either way, that should not be a problem. We have done and seen lots of Garment simulations with higher and lower resolutions.

My subdivisions and iterations are 100 and 120 respectively - To start, is this an adequate amount for the number of verts? 

This is a very high-convergence setup. Which means that if you for example have a high Bend Stiffness on the Cloths, the Cloths will converge pretty well and a high Bend Stiffness (if set really high) might manifest itself in the Cloth looking more like plastic or metal, i.e. very rigid. Similarly, changes to other parameters will also manifest at this high convergence rate. Usually, we have the problem that people don't use enough iterations/subdivisions, so changes to parameters will not have a big visual effect beyond a certain point, as convergence is not reached. You definitely have enough convergence in this case; which is not a bad thing, just be aware that changes to parameters should really give you a good impact and may lead to more rigid looks if you have high stiffness values on the Cloth.
I hope this makes sense. If I had to work with this scene, I would probably reduce the simulation values to maybe 40 iterations and 100 subdivisions, or even lower. At least to start with. You should probably see similar convergence (in practice) and it will simulate quite a bit faster.

Just a quick check, what do you have set for the Simulation Length Scale? Based on the Maya grid by the shoes I would estimate your character to be between 160 and 200 scene units tall, i.e. in centimeters.
In that case, the Simulation Length Scale needs to be set to 0.01.
You probably have set this correctly, but just making sure :)
Getting the Scale right is the first and most important step of any simulation setup.

I hope this helps.

Cheers,
Sebastian
 

Please Log in or Create an account to join the conversation.

Stitch constraints and maintaining drape from Marvelous Designer 1 month 4 weeks ago #796

  • slhouette
  • slhouette's Avatar Topic Author
  • Offline
  • New Member
  • New Member
  • Posts: 6
  • Thank you received: 0
Thanks for the response!! For context, the stitch constraints are to hold the pattern pieces together, they are all separate for each garment. So those stitches are acting as sewing. I'm using the tailored cloth mode. 
So if I understand correctly, in tailored cloth - ie, having flat reference meshes, from the Carbon UVtoPoints - more stiffness/iterations will lead to more convergence to the flat state. That's so interesting! I'll give it a try without using reference meshes. Is there any strategy for maintaining rest length then without the reference meshes? I suppose I'd just have to adjust the stretch stiffness to prevent change in length by stretching/shrinking? 
Thanks for the speedy reply!

Please Log in or Create an account to join the conversation.

Stitch constraints and maintaining drape from Marvelous Designer 1 month 4 weeks ago #797

  • Sebastian
  • Sebastian's Avatar
  • Offline
  • Administrator
  • Administrator
  • Posts: 112
  • Thank you received: 22
Hi slhouette,

So if I understand correctly, in tailored cloth - ie, having flat reference meshes, from the Carbon UVtoPoints - more stiffness/iterations will lead to more convergence to the flat state. That's so interesting! I'll give it a try without using reference meshes

That is correct.

Is there any strategy for maintaining rest length then without the reference meshes? I suppose I'd just have to adjust the stretch stiffness to prevent change in length by stretching/shrinking? 
Thanks for the speedy reply!

Yes, there is :) Good timing. We added a new feature not too long ago, to achieve something like this, which is described in this post:
www.numerion-software.com/index.php/foru...ose-s-folds-wrinkles
So basically, to artificially counter the stretching, let's say in Y direction due to gravity, you can use the Stretch Stiffness Warp value and set it higher, or even reduce the Stretch Equilibrium Warp value to let's say 0.99 to "account" for 1% stretch.
I hope this makes sense.

Cheers,
Sebastian

Please Log in or Create an account to join the conversation.

Stitch constraints and maintaining drape from Marvelous Designer 1 month 4 weeks ago #798

  • slhouette
  • slhouette's Avatar Topic Author
  • Offline
  • New Member
  • New Member
  • Posts: 6
  • Thank you received: 0
Thank you so much for the helpful advice!! If you don't mind I think I have one more question. So I've previously used a UV deformer in houdini (my node setup below) that does the opposite of what Tailored cloth is doing; specifically, increasing my constraint strength holds the shape of the Marvelous Designer draped cloth, rather than trying to reconverge back to flatness. With this Houdini deformer, when I had the problem of pattern pieces flattening out, I strengthened the stiffness of the Weld constraints - the welds sewing the pattern pieces together - and it maintained the shape. I guess what I'm wondering is how the Carbon Tailored Cloth works differently than the deformer setup built in Houdini? 
The UV deformer comes from this video: 

 
Attachments:

Please Log in or Create an account to join the conversation.

Stitch constraints and maintaining drape from Marvelous Designer 1 month 4 weeks ago #799

  • Sebastian
  • Sebastian's Avatar
  • Offline
  • Administrator
  • Administrator
  • Posts: 112
  • Thank you received: 22
Hi slhouette,

There are a few different things going on, which I think we need to clarify; to make sure we are on the same page.
In the video, if I am understanding this correctly, they start with simulated spheres/geometries and use the UVs to map the knitted curves' geometry (which is flat / in XZ plane to the 3D spherical, simulated meshes.
So they already have a simulation and then just perform procedural geometry operations to get the knitted look.
The video really has nothing to do with simulation itself as much as I can tell.

With this Houdini deformer, when I had the problem of pattern pieces flattening out, I strengthened the stiffness of the Weld constraints - the welds sewing the pattern pieces together - and it maintained the shape.

I do not see a deformer in the video, i.e. looks to alread be starting with simulated meshes that are baked out to an alembic.

Let's take a step back...
What does the simulation see when you provide a mesh:
We obviously have the tessellation, i.e. P amount of points, T amount of triangles (or quads) and their connectivity.
But beyond that, a Cloth has soft constraints, i.e. Stretch, Bend, Surface, etc.
For those soft constraints, we need to extract/build our reference data, meaning each edge has a length, each triangle/quad has a surface area, each edge also connects two triangles (or quads) and we can calculate the angle between those two triangles / quads.
This data is what we call the Reference. So when you set Stretch Extension to 1.5, you say that each edge cannot extend more than 50% past the length it had in your reference mesh (or start mesh if you do not provide a separate reference mesh). All subject to convergence, of course.

Normally, Start and Reference Meshes need to have the same tessellation (i.e. same number of points, triangles, connectivity).
Carbon Tailored mode is a specific Cloth Model in Carbon where you have a 3D welded Start Pose, and a Reference Pose that is made from flat panels, i.e. the number of points is not the same.
Internally, and that's the magic of Tailored Cloth Model, we are able to use the metrics from the flat reference panels and apply them to the 3D Cloth mesh.
What that means is that, as you have noticed, all angles in the mesh are flattened (as the reference panels are flat), and all edge lengths / surface areas are extracted from the 2D panels as well.

This gives you two powerful features:
1) You can simulate using the original metrics from Marvelous, i.e. even if your dress stretches to twice the length in Marvelous, Carbon will be able to use the panel data in Tailored mode and drape from there, i.e. we do not apply gravity twice.
2) Having flat references in the XY plane, we now have a space where we can define anisotropy, i.e. Warp, Weft, and Bias.

Just to summarize, Tailored mode doesn't mean that we use some kind of deformer to map the flat meshes onto the 3D Start Mesh / Simulation.
It is a Cloth mode, where the internal Physics engine takes the references (edge lengths, triangle areas) from the flat panels and applies them to the 3D mesh.

Now, to change the topic:
If you are after let's say something like those deformable balls in the video, I would try 2 things:
1) Apply some Surface Pressure to the Cloths. This is a quick and easy way to make something look like an inflated balloon, ball, etc. People have used that for balloons, hair, beards, puffy jackets, etc. 
2) Use a Rigid Welding to stiffen the geometry. We have a few examples in our documentation, like example_tires.hip or example_rubber_cogs.hip (same example are in Maya as well; obvious .mel files).

I hope this helps make things clearer.

Cheers,
Sebastian

Please Log in or Create an account to join the conversation.

Last edit: by Sebastian.
  • Page:
  • 1
Time to create page: 0.282 seconds