PDA

View Full Version : Syflex Crash Test



jhinrichs
04-04-2014, 06:15 PM
http://youtu.be/6mN0n5AgaqM

I got the idea from a Houdini cloth tutorial, forget where I watched that now. It was awhile ago.

Downloaded the character and animation from Mixamo. Both were free for signing up. I can't seem to get Mixamo animations to work in Lightwave though, even following their instructions, so I brought it into Modo and exported a .MDD file. While I was in Modo I created the skirt model.

Back in Lightwave I added the weight map to the skirt model and then applied the Syflex plugin. Had to mess with the setting some. Set the precision down to 0.0001, added gravity, pins (set to soft. Hard pins made the cloth explode), mesh collision, and self-collision.

It did a nice job, but I can't seem to find a self-collision setting that doesn't cause the cloth to clump up when the polygons intersect. You really start to notice it towards the end.

P.S. Rendered in Octane

Crush
04-05-2014, 02:57 AM
This is really a heavy test for Syflex. Perhaps it helps to create a denser mesh on the skirt and/or the character. Can you post a screenshot of the meshes?

Surrealist.
04-05-2014, 05:27 AM
A Friction setting perhaps?

jhinrichs
04-05-2014, 05:44 AM
Here's a couple screen shots.

I agree it's a pretty intense test. Normal movement and a shorter animation probably wouldn't exhibit the clumping problems. The collision seemed to work ok against the girl, the self-collision seems to be where there are issues.

P.S had a little friction on the girl model, that did help some

121187
121188

Surrealist.
04-05-2014, 08:00 AM
I have no idea about Syflex in LightWave. I have only used it in ICE. The friction settings in ICE can be very sticky. So that could be it.

Another thing you might consider is scale. If could be just too small of a scale for good calculations. You might want to up it by 10x or something then export the mdd and model and see how that works.

I don't really see anything that looks too heavy to me. It should be a fairly easy test.

jhinrichs
04-05-2014, 08:20 AM
Hmm, didn't consider scale. Maybe I'll try that

Crush
04-05-2014, 10:43 AM
The density should be ok. Is it better to use double sided cloth with inner and outer polys?

jhinrichs
04-05-2014, 11:10 AM
Not sure on that, the way I read the documentation I infer that it wraps both sides of the polygon with a 'inner' and 'outer' envelope. Not sure if doubling the polygons would help or not.

I tried rescaling the animation this morning, but I'm not really much of animator. I can get the rotations to resize but not sure of an easy way in Modo to rescale all the position keyframes.

jhinrichs
04-05-2014, 11:26 AM
I fixed it, rendering now. Probably take about 4 hours.

I went back and resubdivided the dress model just once this time but made it a subD object, brought it into Lightwave and set Displacement to Last. Messed with the settings for awhile, but other then setting the precision lower, the best settings ended up being the defaults. The simulations were much faster this time.

The only hiccup was around frame 515 a couple of polygons merged and got locked so I manually set the self-collision to zero for a few frames until they unlocked and then set the collision back on.

I'll repost the video when it's fixed.

jhinrichs
04-05-2014, 08:12 PM
Just finished rendering a fixed version of the animation and posted to YouTube. (I deleted the old post on YouTube. I was unclear on how to replace the existing video)

http://youtu.be/1InNMuDD6ak

Turned out much better. Flawless other then I had to clean up about 15 frames here and there in photoshop where there was a knee breaking through the polygons, but otherwise the whole process was pretty smooth. Syflex is sweet software, highly recommended.

realgray
04-05-2014, 10:13 PM
This plugin work with anything else in LW? Wind?

jhinrichs
04-05-2014, 10:16 PM
It's pretty self contained, don't believe it's impacted by native Lightwave dynamics. It does have its own built in wind and air forces though with associated envelopes.

realgray
04-05-2014, 10:21 PM
that's good to hear. I was wondering how you would do things like capes without wind. Glad to hear it's included.

realgray
04-05-2014, 10:38 PM
can you mdd scan the mesh? sorry for asking all the questions but the other thread is really long :)

jasonwestmas
04-05-2014, 10:56 PM
no you cannot mdd scan syCloth

everything that works with syCloth is within syCloth.

Mr Rid
04-06-2014, 03:35 AM
no you cannot mdd scan syCloth

everything that works with syCloth is within syCloth.

?

Of course you can MDD scan SyCloth. You have to in order to get motion blur.

jasonwestmas
04-06-2014, 07:17 AM
?

Of course you can MDD scan SyCloth. You have to in order to get motion blur.

I meant MD Baker doesn't scan SyCloth deformations. SyCloth has its own point cache that you can use, which is inside the deformer.

jasonwestmas
04-06-2014, 08:15 AM
?

Of course you can MDD scan SyCloth. You have to in order to get motion blur.

I just applied clothfx, after syflex, used Scan Motion and saved that as an mdd. Worked perfectly.

realgray
04-06-2014, 10:24 AM
I just applied clothfx, after syflex, used Scan Motion and saved that as an mdd. Worked perfectly.

that's great. Thanks for the heads up.

Surrealist.
04-06-2014, 10:42 AM
Workflow tip:

Don't make the cloth collide with the actual character. Use a proxy object that does not have the same number of polygons. For example for this dress it really only needs to be the legs and they can be slightly larger and smoother even if need be but not even the whole leg.

Calculation will be much faster and you have more control over collision. Hide it when you are ready to render.

Greenlaw
04-06-2014, 10:56 AM
What Surrealist says.

In fact, for coth and hair sims, I usually go a step further by simply parenting spheres and capsules to the bones that approximate the shape of the mesh, and not bothering with MDD for collision (that is, if I can avoid it.) In many cases, the result will be more or less the same thing as when using a fully deforming collision object except the calculations will be significantly faster.

FYI, in Bullet, using non-deforming proxy objects on bones is pretty much necessary because MDD deforming objects for collision is impractically slow.

G.

Surrealist.
04-06-2014, 11:32 AM
I think the issue on the mdd was getting the anim out of Modo.

Greenlaw
04-06-2014, 11:40 AM
Just finished rendering a fixed version of the animation and posted to YouTube. (I deleted the old post on YouTube. I was unclear on how to replace the existing video)
Yeah, YouTube doesn't allow you to replace a video--you need to upload the 'replacement' as a new video. FYI, Vimeo on the other hand, does allow you to replace videos and continue with the comments/hits from the original.

Pretty neat test by the way. :)

G.

Greenlaw
04-06-2014, 11:48 AM
I meant MD Baker doesn't scan SyCloth deformations. SyCloth has its own point cache that you can use, which is inside the deformer.

FWIW, I don't know if it's always been this way or not, but I just now tried MD Baker with SyFlex and it works fine. Just add MD Baker after SyFlex, bake, disable or remove SyFlex, and then add MD Reader and import the MDD.

Using LightWave 11.6.2 and SyFlex 1.3.1.

G.

jasonwestmas
04-06-2014, 01:07 PM
FWIW, I don't know if it's always been this way or not, but I just now tried MD Baker with SyFlex and it works fine. Just add MD Baker after SyFlex, bake, disable or remove SyFlex, and then add MD Reader and import the MDD.

Using LightWave 11.6.2 and SyFlex 1.3.1.

G.

Yeah it's working now, not sure what I did wrong (haven't used LW for ages it seems), could be it was fixed in the latest version. I know that LW is picky about your folder configuration for point cache in some cases.

Greenlaw
04-06-2014, 01:13 PM
Yeah, I don't know either...I haven't used SyFlex for LightWave since 2007. But now that development for LightWave appears to be active again, I'm going to be using it a lot more frequently.

G.