Daz 3D Introduces dForce Physics Engine

1679111238

Comments

  • AllenArtAllenArt Posts: 4,275
    edited October 2017

    If you want buttons or whatnot to move along without melting, assuming they are properly attached, is there an easy setting to keep them solid?

     This. I also want to know this. For future reference.

    Follow Mada's tutorial and paint them with a weight map. Also, Richard mentioned a rigid follow node or something like that...but he didn't explain and I of course have no idea what he meant. LMAO Might be referring again to weight maps like they use for regular conformers rigging.

    Post edited by AllenArt on
  • Wait. WAIT WAIT... somebody. Somebody with a copy that's working! Please test to see if we can somewhat simulate.. soft body dynamics!

    Put the Dforce dynamic modifier on Genesis 3 Female or something. Weight map the entire body so that only her breasts will be dynamic. Turn the dynamic amount to to .3 or something, increase stiffness... or whatever to help 'em keep their shape.

    Because I'm dying to experiment and I can't!!! I need minions!!!!!

  • Oso3DOso3D Posts: 13,400

    I mean, I assume turning down stretching and buckle and so on would work, but curious if there's something like dynamic 0 (but not that because that will act like a pin)

  • AllenArt said:
     

    Follow Mada's tutorial and paint them with a weight map. Also, Richard mentioned a rigid follow node or something like that...but he didn't explain and I of course have no idea what he meant. LMAO Might be referring again to weight maps like they use for regular conformers rigging.

    I think you missed my memo on not being able to follow Mada's tut.

    Anyway, the one time I got it to work, the part I had weight mapped out simply didn't move. And in her tut, she simply had a spot of the shirt that wouldn't move. I would want buttons to move along with a part of the shirt.

     

  • AllenArtAllenArt Posts: 4,275

    I mean, I assume turning down stretching and buckle and so on would work, but curious if there's something like dynamic 0 (but not that because that will act like a pin)

    You can turn the dynamics down to zero...so try .01 maybe and see if that works? LOL

    Laurie

     

  • AllenArtAllenArt Posts: 4,275
    AllenArt said:
     

    Follow Mada's tutorial and paint them with a weight map. Also, Richard mentioned a rigid follow node or something like that...but he didn't explain and I of course have no idea what he meant. LMAO Might be referring again to weight maps like they use for regular conformers rigging.

    I think you missed my memo on not being able to follow Mada's tut.

    Anyway, the one time I got it to work, the part I had weight mapped out simply didn't move. And in her tut, she simply had a spot of the shirt that wouldn't move. I would want buttons to move along with a part of the shirt.

     

    Then rigid follow node seems like it might be the solution. But I can't help you there. Sorry. I've rigged very little in studio.

    Laurie

  • I mean, I assume turning down stretching and buckle and so on would work, but curious if there's something like dynamic 0 (but not that because that will act like a pin)

    If you figure out, please share... I'm sure others will be doing the same but it doesn't hurt to shout more than once. I'm stuck at being just a spectator over here alas.

     

  • For a Rigid Follow Node you don't have the buttons (or whatever) in the base mesh, instead they are separate props (a good candidate for instancing). Use the Geometry Editor tool to select a polygon that the button should attach to, right-click, Geometry Assignment>Create Rigid Follow Node from Selected - this adds a new node as a child of the figure, with an I-beam icon. You cana lso select polygons and asign them to the reference group for the node, they will tell DS how the node should twist etc. (as a unit,). Now parent the buttons or whatever to the Rigid Follow Node(s) and they should move with the mesh without deforming.

  • Oso3DOso3D Posts: 13,400

    Ah, and rigid follow nodes work even in dforce? Excellent

  • veenveen Posts: 38

    Hello we all being busy with Physics i guess;) so here is my second question on this thread: i am playing with the Physics essentials package, and than the guy under the blanket scene.

    So how do i make the blanket thinner if it would look like a satin blanket, its not very detailed with the basic settings.

     

  • veen said:

    Hello we all being busy with Physics i guess;) so here is my second question on this thread: i am playing with the Physics essentials package, and than the guy under the blanket scene.

    So how do i make the blanket thinner if it would look like a satin blanket, its not very detailed with the basic settings.

    If you want finer detail you need to replace the blanket with a higher-resolution plane.

  • veenveen Posts: 38
    veen said:

    Hello we all being busy with Physics i guess;) so here is my second question on this thread: i am playing with the Physics essentials package, and than the guy under the blanket scene.

    So how do i make the blanket thinner if it would look like a satin blanket, its not very detailed with the basic settings.

    If you want finer detail you need to replace the blanket with a higher-resolution plane.

    Thank you very much for helping

     

  • algovincianalgovincian Posts: 1,872
    edited October 2017

    Create -> New dForce Modifier Weight Node is your friend!

    Add an Influence Weights Map and paint the influence of the button to zero, and the fabric surrounding the button so it fades from zero to full influence as it gets further away from the button. 

    I've had limited time to play, but a quick test using Eowyn's overdress worked slick as a whistle:

    Hope this helps.

    - Greg

    button-sim.jpg
    575 x 751 - 128K
    Post edited by algovincian on
  • For a Rigid Follow Node you don't have the buttons (or whatever) in the base mesh, instead they are separate props (a good candidate for instancing). Use the Geometry Editor tool to select a polygon that the button should attach to, right-click, Geometry Assignment>Create Rigid Follow Node from Selected - this adds a new node as a child of the figure, with an I-beam icon. You cana lso select polygons and asign them to the reference group for the node, they will tell DS how the node should twist etc. (as a unit,). Now parent the buttons or whatever to the Rigid Follow Node(s) and they should move with the mesh without deforming.

    How do you parent the buttons? They're usually not even bones on the shirt; just separate surface zones. Man is this a learning experience today!

     

  • algovincianalgovincian Posts: 1,872

    Create -> New dForce Modifier Weight Node is your friend!

    Add an Influence Weights Map and paint the influence of the button to zero, and the fabric surrounding the button so it fades from zero to full influence as it gets further away from the button. 

    I've had limited time to play, but a quick test using Eowyn's overdress worked slick as a whistle:

    Hope this helps.

    - Greg

    Also, I should add that proxies for the full characters speed the sims greatly. In this case, the proxy mesh I created has 6,230 faces, as opposed to the 17,000 that a base resolution G3F has. It is also worth the time investment to learn how to retopologize existing clothing for sim purposes, and then bake the textures to the new UVs of the retopologized mesh.

    Hope this helps (and doesn't add to anyone's confusion).

    - Greg

  • Aave NainenAave Nainen Posts: 1,090

    For a Rigid Follow Node you don't have the buttons (or whatever) in the base mesh, instead they are separate props (a good candidate for instancing). Use the Geometry Editor tool to select a polygon that the button should attach to, right-click, Geometry Assignment>Create Rigid Follow Node from Selected - this adds a new node as a child of the figure, with an I-beam icon. You cana lso select polygons and asign them to the reference group for the node, they will tell DS how the node should twist etc. (as a unit,). Now parent the buttons or whatever to the Rigid Follow Node(s) and they should move with the mesh without deforming.

    How do you parent the buttons? They're usually not even bones on the shirt; just separate surface zones. Man is this a learning experience today!

     

    Just drag them onto the node symbol in scene

     

  •  

    How do you parent the buttons? They're usually not even bones on the shirt; just separate surface zones. Man is this a learning experience today!

     

    Just drag them onto the node symbol in scene

    I think you miss my point. Here I am in the scene, doo dee doo doo... and I have this dress.

    The dress doesn't list "buttons" in it's tree.

    So how can I drag them onto the node symbol in the scene if they're not there?

     

    I did manage to get it to work somewhat. I got up to the parenting part - trying to get a belt not to fall apart - and didn't do any parenting and things stayed together. But the crazy program still keeps crashing.

     

     

  • For a Rigid Follow Node you don't have the buttons (or whatever) in the base mesh, instead they are separate props (a good candidate for instancing). Use the Geometry Editor tool to select a polygon that the button should attach to, right-click, Geometry Assignment>Create Rigid Follow Node from Selected - this adds a new node as a child of the figure, with an I-beam icon. You cana lso select polygons and asign them to the reference group for the node, they will tell DS how the node should twist etc. (as a unit,). Now parent the buttons or whatever to the Rigid Follow Node(s) and they should move with the mesh without deforming.

    How do you parent the buttons? They're usually not even bones on the shirt; just separate surface zones. Man is this a learning experience today!

    You'd need to edit the geometry to remove the existing buttons. This is probably one of the things content creators will want/need to do to make dForce compliant clothing. The trouble with zeroing Dynamic Strength on buttons etc. is that they then pin the cloth in place (assuming they are attached and not floating)

  • veenveen Posts: 38

    maybe some smort guy can collect all knowledge discussed here and make a short youtube movie for all users over here? i have seen Mada De Leeuw her video tutorials and they look very great but i have the idea it doesnt cover everything disussed here.

     

  • marblemarble Posts: 3,453

    After almost 10 pages of this thread a couple of things seem clear to me. One is that dForce is not yet stable enough for general release and I don't have the patience to put up with repeated crashing (the very same reason I returned VWD for a refund but, thankfully, dForce is free). Secondly, I still think that the major feature still lacking is the ability to interactively tug the cloth into place while the simulation is running (something VWD and, more effectvely, Marvelous Designer, have to their advantage). Surely this would negate the obvious difficulties with intersecting meshes and help avoid the crashing.

  • algovincianalgovincian Posts: 1,872
    DAZ_Brian said:

    Other physics simulation types are expected to follow. 

    This is exciting news, @DAZ_Brian!

    Hard data on the sim for a re-topo'ed G2F item Eowyn's overdress (39,913 quads) with a "pinned" button simulated over 100 frames of animation of a G3F proxy figure (6230 quads) getting into position:

    Total simulation time on a modest I5 with 32GB RAM and a 6GB 1060: 2 min 40 secs

    Total VRAM usage: 2372 MB

    Definitely impressed with both the end result, as well as the speed and resources used. @DAZ_Steve did not disappoint!

    - Greg

  • algovincianalgovincian Posts: 1,872
    marble said:

    After almost 10 pages of this thread a couple of things seem clear to me. One is that dForce is not yet stable enough for general release and I don't have the patience to put up with repeated crashing (the very same reason I returned VWD for a refund but, thankfully, dForce is free). Secondly, I still think that the major feature still lacking is the ability to interactively tug the cloth into place while the simulation is running (something VWD and, more effectvely, Marvelous Designer, have to their advantage). Surely this would negate the obvious difficulties with intersecting meshes and help avoid the crashing.

    Both VWD and Marvelous designer are not native, which is a big deal to me (ability to script/automate). Also, they are narrow physics solutions. @DAZ_Brian indicated in the initial post that dForce is going to be used to implement a more robust, unified physics.

    - Greg

  • @marble - I think in my case it's a case of choosing the wrong prop to convert coupled with a memory issue for the most part. Sure there may be other things but it's a learning curve, too.

    Okay, finally did it and got an image done. Attached. Cookie's dress and her hair were both dynamically moved into place by Dforce. So here's what I figured out while fighting this.

    1. Smoothing modifiers actually help with collision issues. The dress refused to take all of Cookie's morphs. I ended up using a smoothing modifier to get it to go around her thunder thighs. That worked well.

    2. If you put a deformer on an object you intend to drape, the item will keep the deformer shaped area intact. I discovered this while trying various deformers in various things to avoid ears and thunder thighs.

    3. Rigid follow node can be useful, but when I tried it for the hair on a whim it created issues. The hair wanted to fly because the follow node wanted to stay or move, what have you. When it moved because the follow node moved, Dforce would actually simulate that flight. I can see doing this for a floating ghost or something like that.

    4. To get the hair to stay in place I simply parented it to Cookie's head and told the underside not to have any dynamic weight.

    5. I need more RAM.

    6. The white belt on the dress is not welded nd originally that portion of the dress fell apart. I selected the bottom line of the skirt top and the bottom line of the shirt top and assigned them as a rigid follow node - I actually did both, make follow node and the one next to that command - and the dress then stayed together. I'm pretty sure it was the second command that did the trick. "Set rigid follow node reference" I think it was. I probably just got lucky but still, there is that.

    7. The wind tunnel has to be directly on the object you want winded. Which means if you want to wind up a whole scene you're gonna have an elbow grease issue, because they've set limits on how big it gets.

    8. I really need more RAM.

    9. And finally. If you set a body part to be dynamic, you can ALMOST get soft body dynamics. The part moves. What it lacks is being able to puff it up. If Dforce had a function that's like Optitex's "internal pressure" setting, I can see doing some nice stuff with that.

     

    This is like.. Christmas in October for me. This Dforce thing. I need more RAM.

     

    cookie.png
    800 x 1035 - 1M
  • 8. I really need more RAM.

    Plenty of places to get it from lol.

  • IvyIvy Posts: 5,315

    This is a game changer for me. I've used VWD extensively for animations, but often get frustrated by it's touchy memory problems,
    I did a quick animation using a Gen 08 figure that I ran through Mixamo ( when mixamo worked right )  A scene like this in VWD would take a few hours. This took me 6 minutes!!!
    The other is Gen 2. The dress doesn't fit very well because this is the fbx from mixamo not a gen 2 figure so it won't autofit correctly. So far, out of the box, I think the results are great. 

    [youtube]

    great job . looks excellent.

  • agent unawaresagent unawares Posts: 3,513
    edited October 2017
    marble said:

    After almost 10 pages of this thread a couple of things seem clear to me. One is that dForce is not yet stable enough for general release and I don't have the patience to put up with repeated crashing

    Same here. I haven't been able to get it to run stably enough to run even half a dozen back to back simulations. It does seem slightly better with less polys in the draped model.

    Also, the cloth simulation in dForce so far looks far too dependent on how many polys a model has. If I use dForce to drape a very well-behaved MD dress, using similar settings, it droops until most of it has fallen right off the figure. If I export a much lower-poly version of the dress, it holds its shape somewhat and stays on the model. That's unacceptable for me in a cloth sim. Yes, a dress with more polys has more points to interact with itself and more places to bend so minor details of the drape should certainly be different. But 4X polys shouldn't mean the dress suddenly stretches 4X as far with (presumably) the same fabric.

    Post edited by agent unawares on
  • marblemarble Posts: 3,453

    @marble - I think in my case it's a case of choosing the wrong prop to convert coupled with a memory issue for the most part. Sure there may be other things but it's a learning curve, too.

    Okay, finally did it and got an image done. Attached. Cookie's dress and her hair were both dynamically moved into place by Dforce. So here's what I figured out while fighting this.

    1. Smoothing modifiers actually help with collision issues. The dress refused to take all of Cookie's morphs. I ended up using a smoothing modifier to get it to go around her thunder thighs. That worked well.

    2. If you put a deformer on an object you intend to drape, the item will keep the deformer shaped area intact. I discovered this while trying various deformers in various things to avoid ears and thunder thighs.

    3. Rigid follow node can be useful, but when I tried it for the hair on a whim it created issues. The hair wanted to fly because the follow node wanted to stay or move, what have you. When it moved because the follow node moved, Dforce would actually simulate that flight. I can see doing this for a floating ghost or something like that.

    4. To get the hair to stay in place I simply parented it to Cookie's head and told the underside not to have any dynamic weight.

    5. I need more RAM.

    6. The white belt on the dress is not welded nd originally that portion of the dress fell apart. I selected the bottom line of the skirt top and the bottom line of the shirt top and assigned them as a rigid follow node - I actually did both, make follow node and the one next to that command - and the dress then stayed together. I'm pretty sure it was the second command that did the trick. "Set rigid follow node reference" I think it was. I probably just got lucky but still, there is that.

    7. The wind tunnel has to be directly on the object you want winded. Which means if you want to wind up a whole scene you're gonna have an elbow grease issue, because they've set limits on how big it gets.

    8. I really need more RAM.

    9. And finally. If you set a body part to be dynamic, you can ALMOST get soft body dynamics. The part moves. What it lacks is being able to puff it up. If Dforce had a function that's like Optitex's "internal pressure" setting, I can see doing some nice stuff with that.

     

    This is like.. Christmas in October for me. This Dforce thing. I need more RAM.

     

    Maybe you will get more RAM for Christmas. :)

    Thanks for your comments. My main concern, not having even downloaded the beta yet, seems to be the mesh intersecting, or body parts passing through each other (like fingers into thighs or, perhaps, thighs into chairs). If something like that (which can be difficult to spot) causes the cloth to explode or the whole program to crash, then there's work to do yet, I think. I guess that people will eventually get used to workarounds like hiding surfaces but I'd rather see these problems handled within the software.

  • algovincianalgovincian Posts: 1,872
    marble said:

    After almost 10 pages of this thread a couple of things seem clear to me. One is that dForce is not yet stable enough for general release and I don't have the patience to put up with repeated crashing

    Same here. I haven't been able to get it to run stably enough to run even half a dozen back to back simulations. It does seem slightly better with less polys in the draped model.

    Also, the cloth simulation in dForce so far looks far too dependent on how many polys a model has. If I use dForce to drape a very well-behaved MD dress, using similar settings, it droops until most of it has fallen right off the figure. If I export a much lower-poly version of the dress, it holds its shape somewhat and stays on the model. That's unacceptable for me in a cloth sim. Yes, a dress with more polys has more points to interact with itself and more places to bend so minor details of the drape should certainly be different. But 4X polys shouldn't mean the dress suddenly stretches 4X as far with (presumably) the same fabric.

    I agree - ideally, more verts would not equal more weight. You can compensate for it by scaling the density, though, no?

    - Greg

  • marble said:
     

    Maybe you will get more RAM for Christmas. :)

    Fat chance. LOL. I'm lucky to get a card... maybe it'll be a RAM shaped card. LOL

     

    marble said:

    Thanks for your comments. My main concern, not having even downloaded the beta yet, seems to be the mesh intersecting, or body parts passing through each other (like fingers into thighs or, perhaps, thighs into chairs). If something like that (which can be difficult to spot) causes the cloth to explode or the whole program to crash, then there's work to do yet, I think. I guess that people will eventually get used to workarounds like hiding surfaces but I'd rather see these problems handled within the software.

    The hair Cookie is wearing I can tell you - from welding it to a mask to make it dynamic for Optitex a long time ago - has a LOT of overlapping areas. It's all overlap. I told it not to collide with itself and it draped beautifully. I was actually a little surprised at how nicely it went.

    I'm gathering that what works is almost hit or miss. But converting to other programs was that way, too. I just mark what works as I go and plug along.

     

Sign In or Register to comment.