Some love for blender cloth simulation

General Forums Blender and CG Discussions
1

While learning cloth simulation in blender i found that the system is really slow compared to others that react instantly as marvelous designer and other research i have see and test. blender already have a powerful set of modeling tools and addons that are good tools for modeling the garment and the only thing that remains is to optimize the cloth system to be more faster, today artist are using marvelous designer and other systems that allow more realistic simulation in clothes wrinkle, folds etc… there is already a sewing feature in blender 2.70 by thesleepless that is really useful the only problems is the slowness and a elastic feature by this same developer in progress.

i make some test whit this character and i have to say i can’t do the simulation it take a lot of time to compute with static meshes imagine with animation…:no:

Kiba_Final_render test.blend.jpg1264×946 175 KB

or maybe i’m doing something wrong or i need a more advanced computer if any one share my same frustration with blender cloth and found a way to solve it without leaving blender please share you’re comment

then i make a search in the net to see some other research and i found this:

and here is the web application realtime:
http://3d.trimirror.com/en/

Another research base on Adaptive Anisotropic Remeshing

2

and here a folding research really useful for neck folding during simulation like in marvelous designer

3

Are you saying you are going to stump up the money required to hire a development team to improve the blender cloth sim up to this standard ?

Or you could contribute to the Gooseberry project which has one of its development goals of improving the cloth sim
http://gooseberry.blender.org/free-software-targets/
http://gooseberry.blender.org/developer-meeting-pipeline-and-tools/

or be patient

Things don’t immediately happen just by wishing it.

4

I think we all want the cloth simulation in blender to be improved faster. Thanks for the links Richard! Didn’t know that Gooseberry was aiming to improve the cloth simulator until now XD - Great news. Hopefully the project works out really well.

  • So yah, be patient. I wish I could contribute but I can’t really :confused:
5

@Richard
i have contribute to blender by buying books and dvd and i want to contribute to gooseberry but right know i’m unemployed i make this post for motivation purpose and pass some information and thanks for the links i dint know either about the cloth improved thanks.

6

That paper airplane demo is pretty neat.

7

However…
It is much cheeper to buy a powerful application than contributing to any part of development of blender.
I would like to directly contribute to lot lot of unfinished parts of blender. I can’t. Sorry.
By example, 600 euros are not enough for contributing to the development of a new multires modifier…
Pretty enough to buy zbrush.
In topic
Cloth simulation is one thread baking. Very slow.

8

maybe this can’t motivate more to optimize the cloth system

9

I don’t think it is possible to motivate developers by showing videos of what other software can do. Money is the best motivator there is but as Michalis said, paying for development is expensive, so expensive that it is cheaper to just buy the commercial software.

10

@NinthJake
Yeah, but I may be crazy and do something about it.
For a certain part of blender development though.
Not for everything around.

11

I vaguely remember there was some kind of FOSS multiphysics solver that was going to get integrated eventually? I think it covered cloth too.

12

Yeah, theres been talk about seeing what Bullet 3 can do for cloth sims, it would certainly make unifying the physics systems easier.

13

Maya has the best cloth simulation I have tried so far.

But the thing is, with Blender, and with any simulation to use, time is best spent learning how to use it. Simulation of cloth has a lot of various technical issues that you have to overcome.

What makes Maya cloth so nice is a lot of these technical issues are solved by adding features designed to solve them by tweaking parameters and/or painting attributes into the cloth.

Blender Cloth works pretty well. You are not going to get anything better unless you move up to Syflex or Maya nCloth.

But no matter what solution you use you have to understand the workflow. They are all about the same. I have written this up in many places, maybe will come up on google searches, my name with Cloth or even user name, Surrealist (lightWave forums) along with cloth simulation or workflow will likely turn up some results.

No doubt you can find improvement here in simulation. But the current solution does work when you understand how to use it.

14

you can simulate cloth like that in blender in…2030

15

There’s not really anything wrong with Blender’s cloth simulation, in fact it is pretty good. The problem lies in the horrible performance and no matter how good you understand the settings you cannot do anything about that.

Just making the cloth simulation multi-threaded would go a long way.
(though I wouldn’t mind having that dynamic tesselation as well ;))

16

Why not being inspired by this for GB, source code available (non gpl) and the research has been supported by Adobe, ATI, Autodesk, Intel, mental images, Microsoft Research, NVIDIA, Side Effects Software, the Walt Disney Company, and Weta Digital :slight_smile:

Sensitive Couture for Interactive Garment Editing and Modeling - and video youtube

demo code can be found in this page, scroll down. link

This other one has focused on collision and applied to cloth, source code available too

http://www.cs.ubc.ca/labs/imager/tr/2012/ExactContinuousCollisionDetection/BEB2012.html

17

I have found this not to be the case in my tests and production scenarios.

And I found I could fix most of the issues by not only understanding the settings, but developing a proper workflow for cloth. Which by the way is also taught in the Syflex portion of the Softimage manual as well as the nCloth docs. The issues that you can not fix in Blender have nothing to do with performance, really. These things crop up because there are missing features.

Performance can be misleading. You can show off a great “performance” in cloth. But the moment you put it to an actual useable simulation situation with a character even the best solutions break down on the performance side.

Of the cloth I have used so far, not one of them has blown me away with performance. With the exception of cloth in Messiah which is stupid fast. Only problem is, it lacks any features that allow you to do anything with it.

And that to me is about the simulation - and features you have to control it - which is why Maya rules. Not because the sims are faster - not by a long shot - but because you have the features to actually control the simulation. That is the key. And in Maya with the nucleus node you actually have to turn up substeps to get the sims to actually work right. You work with lower substeps to improve performance until the final sim… not gonna work. You will be playing with settings that have nothing to do with a feature but are tied into the quality of the simulation which unfortunately kills performance.

In a real production environment it is not unreal to have several minutes or longer per frame just in simulation time. Read some of the articles where studios are doing lots of cloth sims and the resources they use to pull that stuff off is pretty scary. I have found that to be about the appropriate amount of effort computer power-wise to do serious cloth. These guys have farms just dedicated to sims.

You won’t find me arguing that Blender cloth could not be faster.

What is really needed for Blender cloth however is more features that come in handy when simulating character cloth. Features that solve issues you otherwise have to work around.

Of course improving performance is a great help. But that alone will not solve the issues you have to deal with.

If you can find a back issue of 3D Artist Mag, I did a tutorial a year or two ago on Blender cloth. It covers a very basic simulation set up for a character, and offers some solutions.

Here is a list of the tuts I did:

Obviously since then I have moved on to other cloth solutions.

I’d like to see Blender cloth get some love too. :eyebrowlift:

18

The main problem with your example above is the method you are using.

There are two issues that I see which are holding you back.

  1. You are using the actual character mesh - from what I can see.
  2. You have a very highly detailed cloth object, folds buttons and so on. I can’t tell some of that might be just texture. But regardless, you need to simplify your situation.

Have a look at the image down this page a ways.

It shows a collision object that is oversimplified. Used in place of the actual body.

And the cloth object is very straight forward and simple. Usually you simulate on a simple object and use some method to drive a heaver object. Admittedly this can be a little tricky in Blender. But there are options.

There is a lot to this workflow. I am happy to help you further if you post in the support section. I don’t want to get too derailed here.

Send me a message if you make a post and I can show up and give you some further tips.

19

Richard, considering you seem to have a intimate understanding of Cloth Sims, and improved cloth sims are a goal of Gooseberry, have you considered setting up a document where you explain what is good and bad about the current Blender cloth sim to a developer, and what would really help to improve it?

The devs don’t like feature requests(Because often they are poorly thought-out), but they do appreciate help in finding the right design of functionality.

EDIT: It seems Psy-fi’s the person responsible for investigating cloth, and as the texture paint GSoC shows, he’s always very interested in user-feedback, so you could PM him.

20

I think it’s been indicated before by the dev. team that cloth simulation is due for a round of improvements, as even Ton has admitted that trying to simulate it along with hair for Sintel was a disaster.

So it’s going to have to see improvements real soon, as it is probable that the current system would serve as a major roadblock for Gooseberry.