dForce exploding content - possible work around?

2»

Comments

  • D OvD Ov Posts: 103
    edited May 2018

    ...and one other thing that is good about DS - this forum is full of very helpful people. They helped me in the past tremendously.

    In case you are wondering, here's the animation I ahve been trying to apply the DForce to.

    Post edited by D Ov on
  • Richard HaseltineRichard Haseltine Posts: 77,478

    I should think there are issues with the elbow on the knee in the opening, the hands on the hips, and quite likely the mid-thighs clipping with the swaying walk

  • D OvD Ov Posts: 103

     Hey Richard. My guess too. At some point she stands with hands akimbo. So the lesson is - under no circumstances let the hands touch the clothes? I tried the solution offered by Aaave and turn Visible in Simulation off for hands, but then the pants just slide down on the floor altogether...

  • Richard HaseltineRichard Haseltine Posts: 77,478

    The hands can come into contact with the clothes, as long as there is room for the cloth between the hands and the hips (in this case). This isn't just a dForce issue, I recall a 3D World article (a few years back now) on one of the films with a lot of cloth and the makers had a lot of trouble avoiding issues where the cloth got trapped and the sim ran amok.

  • D OvD Ov Posts: 103

    hmm. let me play with it. Thank you.

  • maikdeckermaikdecker Posts: 1,972
    D Ov said:

    Ok, my overall conclusion about this whole DForce business -

    #1 - it's not plug and play.
    #2 - It will not work unless you deal with a DForce taylored items.
    #3 - Or unless you want to spend hours on changing geometry, weight maps etc.. i.e basically remodeling the item. Or if you want the figure to stay still.

    #1 - right you are, it isn't. Except sometimes it is... depends on the piece of clothing you attempt to dForce.

    #2 - some Genesis and Genesis 2 clothes work perfectly "out of the box" with dForce. I've tried quite a few. Yes, there are also some that explode extremely fast or even just "disappear", but on the other hand many clothes work well without having to adjust any values.

    #3 - the more "simple" a piece of clothes is - no bells and whistles, buttons, zippers, decorations, stuff being added for the looks etc. - the higher is the chance that it will dForce without any "remodelling" being needed. And most poses work great, if you just make sure that there ain't no two "static" surfaces influencing the "dynamic" one.

    TLDR: keep it simple. No complicated clothes and no too complicated poses. Although even then there might still be surprises wink

  • D OvD Ov Posts: 103

    Hey maikedecker, thanks fo the comments. My problem is I have been trying to apply DForce to animation, not just still poses. I can't think of think of how I can get say, a shirt tuck under the belt without getting to explode. Also, some suits are meant to collide with shirts e.g. business suit has a shirt, trousersm sometimes a tie or even a vest. The shirts do have buttons, can's help that. So, even if I try, some things are menat to be, I keep banging my head against the wall and experimenting with different clothes. I would certainly change the not working clothes but I wanted to keep the style in the 70s and that sort of outfit is hard to find for G2 models.

  • nicsttnicstt Posts: 11,330

    It works with non-dforce items, but with varying success. Some are pretty straigtforward.

    A frequent culprit is bows, laces and anything else 'stringy'; greatly reduce or turn off dynamic strength may be the only thing that works with them sometimes. It is trial and error.

    Also set layers for different parts can help.

  • D OvD Ov Posts: 103

    Hey nicstt, do you know if layering will work with buttons on shirts? How do I do it?

  • nicsttnicstt Posts: 11,330
    edited June 2018

    Buttons are a perenial problem; usually they aren't attached. So far where the buttons have been vital, I've used VWD to simulate. It has more controllability, but it is more fiddly, and of course costs.

    When there isn't going to be much movement, as in you're just wanting the mesh to relax and gain some 'texturing' - turning buttons off is fine.

    https://www.daz3d.com/forums/discussion/208141/how-to-use-dforce-creating-a-blanket-draping-clothes-on-furniture-and-much-more#latest

    The above thread has lots of useful tips, well worth the time taken to read through it. Don't recall anything that would help with buttons, other than weightmapping perhaps. I've not tried that yet, and it won't give a way to keep buttons parrented to a specific section of geometry.

    You could try reducing the dforce amount; Dynamic Strength. Turn it down from 1, to 0.9, 0.8 or similar. It will affect how much it moves, so a piece of cloth could very well move away too much and become noticeable.

    I have also turned buttons off then just moved them in Blender and sent them back as a temp morph; not ideal but is an option.

    Post edited by nicstt on
  • D OvD Ov Posts: 103

    Thank you nicstt. Very useful stuff.

  • Luv LeeLuv Lee Posts: 220
    mephoria said:

    I attempted to do a systematic study of what settings contributed most to "explosion" of simulated fabrics. Because there were too many complexities in off-the-shelf models, I created my own which included just one potential trouble spot -- a plane with an added toroid which introduced a non-planar topology. (See attachment 1.) As more or less expected, it exploded when I tried to drape it over a genesis figure. To my surprise, however, the problems didn't end up being primarily due to self-collision or the odd topology. I was able to eliminate the problem with various combinations of setting changes:

    1. Simply setting the "bend stiffness" to .25 instead of .5 was, in itself, sufficient to fix the problem. (See attachment 2. The simulation is plausible, but the "donut" section completely collapses, as the fabric lacks stiffness.)
    2. Setting "buckling stiffness" to its minimum was enough to almost fix the problem. (It didn't blow up until the last moment.) Adding in decreases to either stretch stiffness or shear stiffness was enough to make it behave properly.
    3. Setting "dynamics strength" to 90 was enough to keep it from exploding, but also limited the amout of draping. Combinining this reduction with a figure that already has morphs might provide a good hybrid approach for some situations.
    4. Finally I went back to zbrush and actually changed the geometry of the model.Retopologizing so that it had only quads instead of a mix of quads and triangles seemed to be all that was needed. The model simulated smoothly and easily. (See attachment 3. The draping is more interesting and the shape of the donut doesn't get completely lost.)
    5. "Decimated" versions (triangle-based) of the original model exploded in the same way as the original. Decimated versions of the retopologized version didn't explode, though the simulation wasn't as pretty, due to reduced polygon count.

    Summary:

    1. Clean topology matters. It's probably necessary to have non-intersecting "watertight" meshes, but it's not sufficient.
    2. If you have no control over the topology, try simply tweaking the bend stiffness. That may be all you need to change.

     

    YOU ARE MY HERO, like seriously. And yes, I am screaming. LOL

  • odysseyowlodysseyowl Posts: 16
    edited May 2019
    Luv Lee said:
    mephoria said:

    I attempted to do a systematic study of what settings contributed most to "explosion" of simulated fabrics. Because there were too many complexities in off-the-shelf models, I created my own which included just one potential trouble spot -- a plane with an added toroid which introduced a non-planar topology. (See attachment 1.) As more or less expected, it exploded when I tried to drape it over a genesis figure. To my surprise, however, the problems didn't end up being primarily due to self-collision or the odd topology. I was able to eliminate the problem with various combinations of setting changes:

    1. Simply setting the "bend stiffness" to .25 instead of .5 was, in itself, sufficient to fix the problem. (See attachment 2. The simulation is plausible, but the "donut" section completely collapses, as the fabric lacks stiffness.)
    2. Setting "buckling stiffness" to its minimum was enough to almost fix the problem. (It didn't blow up until the last moment.) Adding in decreases to either stretch stiffness or shear stiffness was enough to make it behave properly.
    3. Setting "dynamics strength" to 90 was enough to keep it from exploding, but also limited the amout of draping. Combinining this reduction with a figure that already has morphs might provide a good hybrid approach for some situations.
    4. Finally I went back to zbrush and actually changed the geometry of the model.Retopologizing so that it had only quads instead of a mix of quads and triangles seemed to be all that was needed. The model simulated smoothly and easily. (See attachment 3. The draping is more interesting and the shape of the donut doesn't get completely lost.)
    5. "Decimated" versions (triangle-based) of the original model exploded in the same way as the original. Decimated versions of the retopologized version didn't explode, though the simulation wasn't as pretty, due to reduced polygon count.

    Summary:

    1. Clean topology matters. It's probably necessary to have non-intersecting "watertight" meshes, but it's not sufficient.
    2. If you have no control over the topology, try simply tweaking the bend stiffness. That may be all you need to change.

     

    YOU ARE MY HERO, like seriously. And yes, I am screaming. LOL

     

    Post edited by odysseyowl on
  • odysseyowlodysseyowl Posts: 16
    edited May 2019
    mephoria said:

    I attempted to do a systematic study of what settings contributed most to "explosion" of simulated fabrics. Because there were too many complexities in off-the-shelf models, I created my own which included just one potential trouble spot -- a plane with an added toroid which introduced a non-planar topology. (See attachment 1.) As more or less expected, it exploded when I tried to drape it over a genesis figure. To my surprise, however, the problems didn't end up being primarily due to self-collision or the odd topology. I was able to eliminate the problem with various combinations of setting changes:

    1. Simply setting the "bend stiffness" to .25 instead of .5 was, in itself, sufficient to fix the problem. (See attachment 2. The simulation is plausible, but the "donut" section completely collapses, as the fabric lacks stiffness.)
    2. Setting "buckling stiffness" to its minimum was enough to almost fix the problem. (It didn't blow up until the last moment.) Adding in decreases to either stretch stiffness or shear stiffness was enough to make it behave properly.
    3. Setting "dynamics strength" to 90 was enough to keep it from exploding, but also limited the amout of draping. Combinining this reduction with a figure that already has morphs might provide a good hybrid approach for some situations.
    4. Finally I went back to zbrush and actually changed the geometry of the model.Retopologizing so that it had only quads instead of a mix of quads and triangles seemed to be all that was needed. The model simulated smoothly and easily. (See attachment 3. The draping is more interesting and the shape of the donut doesn't get completely lost.)
    5. "Decimated" versions (triangle-based) of the original model exploded in the same way as the original. Decimated versions of the retopologized version didn't explode, though the simulation wasn't as pretty, due to reduced polygon count.

    Summary:

    1. Clean topology matters. It's probably necessary to have non-intersecting "watertight" meshes, but it's not sufficient.
    2. If you have no control over the topology, try simply tweaking the bend stiffness. That may be all you need to change.

     

    I narrowed down the real issues with dForce. It the hip! The hip Y or elevation value must be at zero, any number in the hip will explode the clothes. See attached. See Surface parameter value. It explodes when figure is moved up or down or away from ground zero. It explodes at default negative value on hip which is relative to ground zero. It DOES NOT explode when value is zero on hip anywhere above or below ground zero. Duplicate my experiment and let me know the results? 

    dforce_on-plane.jpg
    3840 x 1200 - 1M
    dforce_hip-at-zero.jpg
    3840 x 1200 - 1M
    dforce_move-above-plane.jpg
    3840 x 1200 - 2M
    Post edited by odysseyowl on
  • Luv LeeLuv Lee Posts: 220
    mephoria said:

    I attempted to do a systematic study of what settings contributed most to "explosion" of simulated fabrics. Because there were too many complexities in off-the-shelf models, I created my own which included just one potential trouble spot -- a plane with an added toroid which introduced a non-planar topology. (See attachment 1.) As more or less expected, it exploded when I tried to drape it over a genesis figure. To my surprise, however, the problems didn't end up being primarily due to self-collision or the odd topology. I was able to eliminate the problem with various combinations of setting changes:

    1. Simply setting the "bend stiffness" to .25 instead of .5 was, in itself, sufficient to fix the problem. (See attachment 2. The simulation is plausible, but the "donut" section completely collapses, as the fabric lacks stiffness.)
    2. Setting "buckling stiffness" to its minimum was enough to almost fix the problem. (It didn't blow up until the last moment.) Adding in decreases to either stretch stiffness or shear stiffness was enough to make it behave properly.
    3. Setting "dynamics strength" to 90 was enough to keep it from exploding, but also limited the amout of draping. Combinining this reduction with a figure that already has morphs might provide a good hybrid approach for some situations.
    4. Finally I went back to zbrush and actually changed the geometry of the model.Retopologizing so that it had only quads instead of a mix of quads and triangles seemed to be all that was needed. The model simulated smoothly and easily. (See attachment 3. The draping is more interesting and the shape of the donut doesn't get completely lost.)
    5. "Decimated" versions (triangle-based) of the original model exploded in the same way as the original. Decimated versions of the retopologized version didn't explode, though the simulation wasn't as pretty, due to reduced polygon count.

    Summary:

    1. Clean topology matters. It's probably necessary to have non-intersecting "watertight" meshes, but it's not sufficient.
    2. If you have no control over the topology, try simply tweaking the bend stiffness. That may be all you need to change.

     

    I narrowed down the real issues with dForce. It the hip! The hip Y or elevation value must be at zero, any number in the hip will explode the clothes. See attached. See Surface parameter value. It explodes when figure is moved up or down or away from ground zero. It explodes at default negative value on hip which is relative to ground zero. It DOES NOT explode when value is zero on hip anywhere above or below ground zero. Duplicate my experiment and let me know the results? 

    Going to experiment with this as well -- thank you so much.

  • EveniosEvenios Posts: 119

    i just want to say this post really helped me out.  the bend option should really be 0.25 or 0.30 also works by default..........dforce isnt perfect and the more complex the set is the more issues but it can work in time!

  • dijituldijitul Posts: 145

    Most of my problems with exploding dforce cloth come from two things:  1) The collision offset is set to high for the starting position, or 2) the speed of the animation is too fast.  Physics apply here.  If you attempt to animate from the default T-pose to another pose in 30 frames, you're likely going to have a bad time.  That's not how fast people or objects move in the real world.  

    Thus, I recommend adding these two tips to everything you try:  1) Lower collision offset if you see a lot of "errors" fly by when you begin the simulation, and 2) add more frames.  Make your animation cover a realistic amount of time:   30 fps * 6 seconds = 180 frames.  You can't expect 30 frames to resolve properly all the time.

    Hope this helps.

  • Is it normal that it takes almost 20 minutes for one frame (one pose movement for example stand to sit)? 

    System: 2700x / RTX2080ti 

  • I'm unclear if this post is meant to cover "dforce" clothing, or just dynamic cloth, they don't seem to be the same thing. I have a scene I'm working on with a figure wearing two pieces of dforce clothing, but nothing shows up in the Dynamic Clothing tab. I can select one of them and run a simulation, but I don't seem to have any control over the dynamics at all. Sometimes the item will explode, seemingly when it tries to collide with an object that isn't the figure it's attached to. I wish there was a way to add objects to the simulation like you can with dynamic cloth.

  • Steel Rat said:

    I'm unclear if this post is meant to cover "dforce" clothing, or just dynamic cloth, they don't seem to be the same thing. I have a scene I'm working on with a figure wearing two pieces of dforce clothing, but nothing shows up in the Dynamic Clothing tab. I can select one of them and run a simulation, but I don't seem to have any control over the dynamics at all. Sometimes the item will explode, seemingly when it tries to collide with an object that isn't the figure it's attached to. I wish there was a way to add objects to the simulation like you can with dynamic cloth.

    The Dynamic Clothign pane is the old OptiTex system, dForce is controlled through the Simulation Settings pane - with the cloth properties set in the Surfaces pane.

  • Simply setting the "bend stiffness" to .25 instead of .5 was, in itself, sufficient to fix the problem. (See attachment 2. The simulation is plausible, but the "donut" section completely collapses, as the fabric lacks stiffness.)

     

    Thanks Mephoria! That made a huge difference

  • GroveFastGroveFast Posts: 19

    This works even today thanks a lot

Sign In or Register to comment.