Best approach for cloth ?

[sebkaine]

hi guys

 

i would like to know what is the best approach to get

- fast

- simple

- accurate

Cloth in H

 

i need to generate nice smooth silk ribbon that float in air with control of the motion.

 

 

i have try the doc exemple and for the moment i find the FEM solver

- pretty slow

- for a pretty bad outpout

 

I'm even wondering if it would not be wiser to do it in nCloth.

My problem with nCloth is that i loose proceduralism for fast retakes.

 

So i would like to know if you have good tricks for this kind of fx , and what would be the most efficient solver.

I would also like to be sure that my incompetence is the problem and not the solver in itself

 

well thanks for any useful input on the subject !

 

cheers

 

E

 

[Skybar]

I haven't tried it myself, but maybe give the new Grain solver a shot.

[sebkaine]

Thanks David ! In fact the bacon slice exemple definitly has some appeal ... i'm gonna check that !

 

I've also think about some tricks with wire solver which is aces.

 

I would also be curious to know what solver would be the best to make soft bodies stuff , like the old Maya soft bodies engine.

Basically

- each point of your geo is a particle with no springs connection

- and you can if you want add springs connection manually beetween each point.

While old school i find it useful ...  and very fast for sheets and ribbon stuff.

[sebkaine]

Well the problem with granular is that it generate point collision position that are not base on the point of your geo.

 

The workflow i am trying to find is :

- each point of my poly object is a particle

- each particle is link to other with springs

- each point can colide with himself or with other object

- you can control the springs properties and collide properties in a easy way.

 

basically  i need nCloth in houdini. But from my test the FEM while not that bad is really

- slower

- less accurate

- longuer to setup

 

I have not so much time so i guess i'll go nCloth on this one.

 

But thanks !

Edited April 3, 2015 by sebkaine

[Skybar]

Well, it is Houdini after all - you can do pretty much whatever you want. This is the first time I'm doing anything with the new Granular stuff mind you - but I thought I'd give it a quick try to make an example using only the input points. I guess it works quite fine. This is just 5mins work as well, might be better ways to do it.

 

grainCloth_dv.hipnc

Edited April 3, 2015 by Skybar

[sebkaine]

Thanks a lot David this looks extremely promising ! I am actually cooking a scene with FEM cloth , i  change my mind

because while slower i think it's good enough and having the ability to keep the stack 100% updatable is balancing the cons.

 

With Maya you have to start over your connections again and again ... i must be strong and stay away from the middle age workflow

 

I'll post a hip when i'll have something acceptable ! I will also play with your scene cause it look faster than FEM !

 

THX again !

Edited April 3, 2015 by sebkaine

[Guest tar]

Always keep H12.5 around for it's 2x faster non-Fem cloth solver.

[sebkaine]

well cloth is not bad after all, it's definitly usable !

i am trying to understand very basic constrain.

 

but sbd spring constrain is not working as i expected.

 

basically i want to setup a spring beetween two point and be able to play with the spring stiffness and damping.

it looks that the connection is working but the spring has no effects.

 

here is a very basic scene that illustrate the pb , if you have any idea that would be great !

 

 

spring_constrain.hipnc

[SamHowell]

Hello

 

This all sounds quite familar. I gave H13 cloth a go last year for the first time and ended giving up as it was simply too slow. It gave really nice results but the turnaround was just too slow for production (at least in TV). I'm now giving it another whirl with H14 and I am also looking for some silk fine detail cloth floating in the air. There has definately been a speed improvement but a colleague sitting near me using Ncloth has managed to do a very decent job of it in a fraction of the time. Still, I'm not giving up just yet, the procedural constraints are very usefull. So far just turned down the shear, thickness and stiffness right down and have some nice silky results pretty quickly. It grinds to a halt once I upres the mesh but the results seem very natural, especially the self collision.

[sebkaine]

Hello Sam,

 

I have keep using Houdini cloth instead of nCloth. I'm glad i do so , cause while nCloth is still faster, houdini cloth is now perfectly usable in production.

But thing start to really be interesting when you play with custom attributes. The fact that you keep the stack updatable is also a big plus.

 

It would be nice to have some presets.

 

Those 2 pages are quite interesting to setup presets and custom @

http://www.sidefx.com/docs/houdini14.0/cloth/changing_cloth_look

http://www.sidefx.com/docs/houdini14.0/nodes/dop/clothobject

 

playing with those is really powerfull to control things.

@targetP

@targetv 

@targetstiffness

@targetdamping

@pintoanimation

 

i also solve my problem of SBD spring contsrain. i use cloth attach in soft mode instead !

 

Cheers

 

E

Edited April 30, 2015 by sebkaine

[SamHowell]

Thanks for the information sebkaine. I havn't even begun to touch cloth attributes, just using the solver off the shelf to begin with. I'll start digging deeper, I'm not going to give up this time   . The flexibility of the Cloth Attach Constraint is what has made me keep going. it's so usefull.

Edited April 30, 2015 by SamHowell

[sebkaine]

I've just find thanks to a hip posted by Artem that wire solver can be use as a cloth / soft bodie tools,

as we were enumarating FEM cloth alternative in H for faster result here are now all the alternative i see for the moment

 

- SOP spring

- Wire Solver

- POP Granular Solver

 

i post a wire exemple if someone is curious !

cloth_wire.hip

[mirHadi]

Emmanuel! thank you for your file...

 

I don't see the cloth solver, are these methods way better than cloth solver?

It would be great if you can share your experience with cloth sims in Houdini.

 

in fact I used to do these kind of stuff in Maya, I don't like it at all ,and I do want to work it out in Houdini. 

[sebkaine]

@haaranoos.

 

No the wire solver is not a better method imo, it's more a hack / cheats. If you want to do serious cloth stuff like nCloth i would use the FEM cloth solver.

here is a trivial hip

http://forums.odforce.net/topic/22766-why-cloth-sime-so-slow-with-finite-element-solver/?hl=cloth

 

I haven't test cloth in H15, but in H14 it does work fine and it's very usable in production.

the + are :

- working with attribute

- connect to other solver

- stay at 100% in H

 

the - are :

- no presets painful to find the good stifness value.

- slow compare to nCloth if you compute self collision

 

I would still consider nCloth / nHair a better tool at the moment. But the big plus of houdini is that you keep the procedural stack.

But to setup very complex cloth shots or to setup precise material presets with short deadline, i think i will stick with Maya.

The fact that you now have mudbox tool in maya  to setup your blendshape is also a killer tool to correct sim very fast.

 

I haven't test H15 and cloth is far better so test both and make your decision on your own test !

 

i also post this one to show you the potential of H in this field. Make this in maya is just not possible imo

 

Cheers

 

E

Edited December 5, 2015 by sebkaine

[tricecold]

well cloth is not bad after all, it's definitly usable !

i am trying to understand very basic constrain.

 

but sbd spring constrain is not working as i expected.

 

basically i want to setup a spring beetween two point and be able to play with the spring stiffness and damping.

it looks that the connection is working but the spring has no effects.

 

here is a very basic scene that illustrate the pb , if you have any idea that would be great !

 

 

Hey Seb; thnx for the file;

 

I have the same issue, I cant get the constarints moving the cloth object from wires,

I attached my example file.

 

any ideas ?

FEMConst.hipnc

Edited December 26, 2015 by tricecold

[sebkaine]

Him Tim Sorry for the late answer, if you are on 14 SBD spring contsrain doesn't work you had to use an other one, which is better at the end. Let me some time to dive in the mess of my old hip

i have a scene with lot of simple test with cloth constrain i will post it here.

 

But i'm quite sure it's clothAttachConstrain in soft mode

 

I have attach a doc exemple modify in soft mode.

You might also have to check the doc on the applyRelationship node to build lot of constrain in your scene

 

Keep in mind that all the properties can be control throw attributes procedurally which make them very powerful.

 

Cheers

 

E

doc_exemple.hipnc

Edited December 30, 2015 by sebkaine

[MENOZ]

I would like to pop in the thread just to link this very interesting  piece of techology.

http://www.numerion-software.com/carbon-houdini-plug-ins/carbon-cloth

 

SESI WE NEED THIS  

 

Houdini fem cloth is still slow when dealing with high polycount (I don't find turning off self collision an option). 

Sebkaine, I cannot agree with you when you say " it's very usable in production". I guess it depends on the type of "production" ( I prefer to use the word "work" instead of "production", as I find often it's used as a fancy word.)

 

Houdini pbd is much faster but it still lacks the bend resistance parameters, which makes it extremely hard to control the stiffness.

[sebkaine]

I agree that having both :

- a particle based + string cloth like nCloth and

- the FEM cloth would be great

 

I don't know how FEM work under the hood but i must admit it is really slower than nucleus.

 

What i mainly miss from nCloth is all the presets you have, you just pick one and you don't have to search for the good stiffness value.

I think having those presets would be as important as speed. Cause if you batch your sim on your farm , speed is not a fatal issue imo.

 

The fact that you can control all your attributes throw procedural stack is really better in H compare to Maya.

The fact that you can connect your cloth solver to any other solver also open better possibility than Maya.

 

Cheers

 

E

[Ali Haider]

Hi,

 

So, It seems like I am in the exact same position with the similar issue, trying to contain a shape of geometry with very high stiffness and its not working even after giving a crazy high values to constraint iteration + substeps + stiffness , and after reading the above posts, I feel like its the same bend resistance attribute (that we dont have I assume), is there any way to control it?

 

 

grain_stiffness_test.hip

[ghoest]

the carbon solvers are great. I feel like FEM is really great as well but to be functional it has a steep learning curve (steeper than the wire solver). A lot of both fem and wire solver material issues boil down to stiffness ratios. FEM differs from nCloth because it's thing shell tetrahedrons and doesn't use position based dynamics which while quick are not accurate.

The grain solver probly doesn't have bend stiffnesses because there is no concept of bending between two constrained particles. I haven't looked at it but maybe it has an angle constraint that can act like bend resistance?