Animated Dynamic Clothing Technique (proof of concept)

1568101117

Comments

  • JonstarkJonstark Posts: 2,738
    edited December 1969

    My latest attempts were all to attach different vertex soft bodies to different bones, but when they would bump into each other during movement of the character they would explode or at least tear themselves up very badly, and even invisible as they might have been, this was not kind to the cloth that was overlaying them :)

    This approach of a single vertex object that is simply broken into pieces is brilliant, and solves the problem I was running into in a single stroke. I'm actually at work currently, but will definitely be playing with this a bit more, later today.

  • DiomedeDiomede Posts: 15,168
    edited December 1969

    Jonstark said:
    My latest attempts were all to attach different vertex soft bodies to different bones, but when they would bump into each other during movement of the character they would explode or at least tear themselves up very badly, and even invisible as they might have been, this was not kind to the cloth that was overlaying them :)

    This approach of a single vertex object that is simply broken into pieces is brilliant, and solves the problem I was running into in a single stroke. I'm actually at work currently, but will definitely be playing with this a bit more, later today.

    I think some of the explosion issue can be addressed by making sure self-collision is turned off and by checking the margin setting. My Aiko3 long dress example exploded the first time but was very stable after I made sure self collision was turned off for the underarmor and its margin setting. Since I did both at the same time, I'm not positive which was most important. Oops.

  • DiomedeDiomede Posts: 15,168
    edited February 2015

    I realize there are a lot of posts to go through (which is good), but I've tred to gather some of the key ones that discuss parameter settings and similar specifics. Might help people catch up. Just a couple of notes on the underarmor.

    - Stringtheory has a nice overview of the concept with a screenshot of the soft body and soft body attach menus.
    http://www.daz3d.com/forums/discussion/49954/P15/#732510

    - Stringtheory demonstrates an animation showing a loose flowing skirt reacting when a single limb is moved here.
    http://www.daz3d.com/forums/discussion/49954/P45/#733596

    - Stringtheory provides one example of numerical settings for underarmor soft body. It includes stiffness, bending, self collision (note, I turn it off), and for the softbody. It also includes simulation accuracy and geometric fidelity settings in the scene tab. See here
    http://www.daz3d.com/forums/discussion/49954/P45/#733622

    - Stringtheory has a discussion of adjusting settings such as stiffness and bending here
    http://www.daz3d.com/forums/discussion/49954/P60/#734245

    - Marcus Severus discusses alternative settings for some of the parameters, and also some tips regarding the mesh of the draping object, here.
    http://www.daz3d.com/forums/discussion/49954/P60/#734248

    - Marcus Severus has a link to another thread (earlier) on softbody physics settings here.
    http://www.daz3d.com/forums/discussion/37372
    - with video example here
    https://www.youtube.com/watch?v=7radMxgLA8Q&feature=youtu.be
    - along with screenshots of the physics settings
    http://www.daz3d.com/forums/discussion/37372/#550605

    - a link to a PhilW low poly mesh for V4 that might be adapted for underarmor
    http://www.daz3d.com/forums/discussion/49954/P150/#752578

    - Dart with a screenshot of using select by : name : polygon
    http://www.daz3d.com/forums/discussion/49954/P165/#752631

    - Stringtheory with further updates and demonstrations regarding settings and much more
    http://www.daz3d.com/forums/discussion/49954/P195/#753927

    - Me with a screenshot of mutliple softbody attach areas for a single underarmor object dividied into separate bodypart meshes.
    http://www.daz3d.com/forums/discussion/49954/P195/#758041

    I'm not sure the gaps in the underarmor are necessary, but I think it is a good idea to have the underarmor in a single object. It might be convenient to have different meshes within the underamor, even if not necessary, because it is important to assign separate soft body attaches for each bone of the underlying figure. The separate meshes make selection and naming easier.

    I have found that for me, it is easiest to match the underarmor parts to the figure bones if I have first named the vertexes of each portion of the underarmor to match its corresponding bone.

    For example, if in the vertex modeler, I select the entire panel that corresponds to the right thigh, I then use Selection : Name : Vertices and then choose a name like Thigh Right. Then when I do the softbody attaches for the underarmor, it is relatively easy (but tedious) to use the vertex selection tab in the softbody tab rather than the paint tools. Make sure vertices are not assigned to more than one bone. Also make sure that you validate each softbody attach.

    As PhilW has pointed out, you don't necessarily need to model your own underarmor panels because you can always just use the obj for a low poly version of the figure (or the figure itself). Dart shows how to select by polygon group,and these figure objs already have the bodyparts grouped.

    Hope this update helps bring more people up to speed and ready to experiment.

    Post edited by Diomede on
  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    diomede64 said:
    I realize there are a lot of posts to go through (which is good), but I've tred to gather some of the key ones that discuss parameter settings and similar specifics. Might help people catch up. Just a couple of notes on the underarmor.

    And I believe I posted a detailed, step by step procedure for making the underarmor...and some videos showing examples of results...and a detailed tutorial on Bullet soft and rigid physics, including detailed settings recommendations....and a bunch of other stuff....

    Sorry if none of it was helpful.

  • evilproducerevilproducer Posts: 9,050
    edited December 1969

    I'm sure it was a simple oversight.

  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    I'm sure it was a simple oversight.

    It usually is... :)

    But I'm glad we're getting closer to a workaround for this issue. It's unfortunate that we've been left hanging like this with such an unfinished yet important feature.

  • DiomedeDiomede Posts: 15,168
    edited December 1969

    No offense intended. Quite the opposite. I think I have posted 33 or 34 times in this thread, and only included 1 of my own even though I have many more with screen shots and parameter settings. The theme of my posts (repeated several times) is that I value more people posting, more people testing aspects that I am not testing because we all learn from that, and that I value both successful tests and unsuccessful tests. How someone could interpret that as not valuing more posts is beyond me.

    The list is meant merely to have enough information for someone new to the thread to get an overview of the underarmor approach, the role of softbody attach, sample settings, and some explanations of how to construct underarmor. I think the list accomplishes that but reasonable people can disagree. If I didn't omit anything, it wouldn't be a shorter summary to get more people involved.

    Repeat, no offense intended. Familiar stolen pic attached.

    group_hug400.jpg
    400 x 415 - 15K
  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    Anyway, I'd like to say that this image was done using Bullet cloth, but sadly that ain't the case. In fact it's a very nicely designed conforming clothing. And I'd love to get to the point where you can do something like this in Bullet, even with a workaround.

    I know Marvelous Designer can handle something like this no problem. But even without the frills and stuff, it would be great if we could come up with a Bullet cloth sim that could drape and animate something like this. :) :)

    cloth.jpg
    1000 x 729 - 52K
  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    And while this *IS* a Bullet cloth drape, and something I'd like to strive for in a workaround, it's not an animated one. Perhaps I can again get off my butt and see if I can use the armor on something like this and do some animation...

    clothTest_009.jpg
    1200 x 1288 - 415K
  • DiomedeDiomede Posts: 15,168
    edited February 2015

    Anyway, I'd like to say that this image was done using Bullet cloth, but sadly that ain't the case. In fact it's a very nicely designed conforming clothing. And I'd love to get to the point where you can do something like this in Bullet, even with a workaround.

    I know Marvelous Designer can handle something like this no problem. But even without the frills and stuff, it would be great if we could come up with a Bullet cloth sim that could drape and animate something like this. :) :)

    Agreed. Very nice. Well done. I hope the end result will get closer to your pic. I encourage everyone who is interested to keep trying and keep reporting. If you have Marvelous Designer, I would be the last person to tell you not to use that, or Blender, or whatever works. Currently, I use Poser as a cloth plugin, not Carrara at all costs.

    I get very frustrated with conforming dresses, tunics, professorial robes, etc. The attached pic of V4 in the standard morphing fantasy dress in a simple sitting pose illustrates my frustration with conforming clothes in that situation. That is why I use Poser as a cloth plugin.

    So, for anyone interested, lets see what the softbody physics clothes draped on softbody attached underarmor can achieve. If it is no good, I'll go back to using Poser as a plugin and other people can use Marvelous Designer, Blender, or whatever. I think some of the tests have shown real promise, but we won't know until we try. My Aiko3 sitting example, while not stylish or realistic, is leaps and bounds better than the V4 sitting in a MFD conforming figure.

    V4_in_MFD_and_fairytale_hair.JPG
    1068 x 849 - 108K
    Post edited by Diomede on
  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    diomede64 said:
    The attached pic of V4 in the standard morphing fantasy dress in a simple sitting pose illustrates my frustration with conforming clothes in that situation. That is why I use Poser as a cloth plugin..

    And don't forget, if you're doing just a single image, and not an animation, for that morphing fantasy dress you might be able to use Bullet to drape that on a static figure, assuming the dress mesh is in decent shape...and even though it's conforming clothes

  • DiomedeDiomede Posts: 15,168
    edited December 1969

    Yes, good point. At that point I prefer to drape using Poser. Faster, less hassle, and I'm more familiar with the tools. However, this softbody on softbody technique has really accelerated the calculation time in Carrara and made it a much more viable option for me.

    - Genesis 2 Female customized clothing-

    This might be a test that could get some attention because Daz cares about genesis. Plus, you don't have to be a modeler. Just export and then import the obj (may have to do scale adjustments and translate, but not adding or adjusting vertices). Wide gaps between underarmor parts are not necessary, although I do think they should not be welded in a single contiguous mesh. It is OK if they end up intersecting as long as self collision is not checked for the underarmor.

    I loaded the genesis 2 female in Carrara and dialed up a custom shape. (25% each of Aiko6, The Girl 6, Stephanie 6, and Victoria 6). I didn't really care about the shape, other than I wanted it to be a custom shape. I then exported the custom figure as an obj file making sure to include morphs and skinning. I then imported the obj making sure to include under "grouping" to check make one object per group. This brought back to carrara a vertex obj with the same shape as the custom genesis 2 female with a separate mesh for each bodypart. It was the wrong size, which was weird, but easy to correct.

    - constructing the underarmor

    I inserted a new vertex object. To populate it, I alternated jumping in to each imported bodypart group, selecting all, copying, then jumping back to the underarmor vertex object and pasting. The bodyparts would not be welded, which is good. The newly pasted bodypart would still be selected, so I used selection : name : vertices to create the groups. The result is a vertex object made up of many meshes that exactly match the bodyparts of the G2F but are not welded. (after resizing to overlap the G2F). You can then delete the imported figure obj.

    - softbody attach the underarmor

    Select the underarmor vertex obj. Assign a general soft body modifier. Then, assign an additional softbody modifier for each named vertices group (corresponding to bodyparts).

    In the attached, the skirt is yucky, but it should be clear that the legs moved the pleated skirt with no problem. The calculations took less than three minutes.

    genesis_customized_drape_2.JPG
    676 x 611 - 47K
    genesis_customized_drape.JPG
    884 x 805 - 104K
    custom_shape_genesis_custom_underarmor.JPG
    834 x 872 - 82K
  • DiomedeDiomede Posts: 15,168
    edited February 2015

    RE: custom Genesis 2 character underarmor

    In my last post, I mentioned that you don't need modeling skills in order to get a set of bodypart meshes to build your underarmor from. Select the G2F figure. Use FILE : EXPORT. Make sure to check the box to include morphs and skinning. When you import it back, under the group setting, make sure to check Create One Object Per Group. Note - I am forgetting a setting because the size is wrong, but easily corrected.

    The obj will import as a group made up of a vertex object for each bodypart. From here, how do you get a single vertex object made up of them all, but not welded? Insert a new empty vertex object, then copy and paste each mesh into the new vertex object. Name the vertices for each part and it will be easier to complete the soft body attach steps.

    You will likely want to adjust the X and Y scales of many of the underarmor parts so there is less likely to be poke through.

    Conceptually, this works, people. :cheese: :coolsmile: :-)
    Now to refine it with better fitting underarmor for various figures and of course much better actual clothes.

    zzz_import_create_one_object_per_group_2.jpg
    1200 x 774 - 147K
    zzz_export_settings_morphs_and_skinning_2.jpg
    983 x 749 - 131K
    Post edited by Diomede on
  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    diomede64 said:
    Yes, good point. At that point I prefer to drape using Poser. Faster, less hassle, and I'm more familiar with the tools.

    I'm curious...how do you get your Carrara scene into Poser to do the drape?

  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited February 2015

    BTW, speaking of Marvelous Designer....

    From a thread I started about 2-1/2 years ago, July 2012, which I started out by saying "Since Bullet is pretty much dead in the water for the forseeable future..." :) :) ....

    Here's an image of some clothing developed very quickly in Marvelous Designer, which at the time was quite new, and the guy doing it (me) had just started using it. So apparently it was idiot proof.

    The wonderful thing about it is that you design and build your clothes using a method very close to what real clothing makers use....you lay out the pieces on a flat, 2D pattern, and it automatically assembles the pieces at the seams and does an excellent, real time shrink fit. And the UV maps are wonderful, matching the patterns the way you want them too, and the way they do with real clothes.

    Stuff like the gown thing with all the ruffles shown below can be done in minutes. And if you know what you're doing (which I don't when it comes to clothing design), I'm sure you can make stuff that is incredible. And I'm sure it has been vastly improved over what they had in 2012.

    I dunno, it just seems like a nice dream for the future of Bullet....

    Sorry for the diversion into dreamland...

    Gown_010.jpg
    801 x 928 - 29K
    Post edited by JoeMamma2000 on
  • RoygeeRoygee Posts: 2,247
    edited December 1969

    @diomede - that is starting to look really viable :)

    Couple of questions, if I may. The imported .obj will have the same mesh density as the original - so it appears that it is not the mesh density that is the problem with collision, but "something else"? Any idea what that something else may be?

    Secondly - just a suggestion. If you export the .obj, then import as a single .obj, delete the parts not needed, name vertex groups, export as .obj and import as .obj as separate groups, would that not cut down on all that copy/pasting time?

  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    Roygee said:
    Secondly - just a suggestion. If you export the .obj, then import as a single .obj, delete the parts not needed, name vertex groups, export as .obj and import as .obj as separate groups, would that not cut down on all that copy/pasting time?

    I wrote a procedure for importing the .obj as separate parts right on top of the actor/avatar/rigged V4/whatever you call it. So no need to copy and paste anything. All the parts are named like the rigged character.

    Or maybe I'm not understanding his procedure...

  • JonstarkJonstark Posts: 2,738
    edited December 1969

    Work and a long day of family stuff got in the way, now I get back to my computer and it's too late to play, though I started following the steps and assembling an underarmor for a V4, and now have vertex object with separate meshes for forearms, shoulders, collars, thighs, shins, neck, chest, abdomen, and hip. It's really too late for me though and I have to get up early, so my playing will have to wait until after work tomorrow (maybe even *during* work if I get a few slow periods) to experiment.

    Thank you very much for detailing your method in such a step by step Diomede, your testing looks quite good, and this looks very viable and fun to play with! :)

  • JonstarkJonstark Posts: 2,738
    edited December 1969

    Roygee said:
    Secondly - just a suggestion. If you export the .obj, then import as a single .obj, delete the parts not needed, name vertex groups, export as .obj and import as .obj as separate groups, would that not cut down on all that copy/pasting time?

    I wrote a procedure for importing the .obj as separate parts right on top of the actor/avatar/rigged V4/whatever you call it. So no need to copy and paste anything. All the parts are named like the rigged character.

    Or maybe I'm not understanding his procedure...

    I think it's essential that they all be part of the same mesh object, just with detached pieces that can be attached to different parts of the characters. On the other hand, thinking back, maybe your procedure gives that same result, I may be misremembering, and I'll have to dig up the other thread (or was it earlier in this one?) where you mentioned this procedure, because maybe that's an easier/faster way to get to the same result.

  • JonstarkJonstark Posts: 2,738
    edited December 1969

    Roygee said:

    Couple of questions, if I may. The imported .obj will have the same mesh density as the original - so it appears that it is not the mesh density that is the problem with collision, but "something else"? Any idea what that something else may be?

    Joe mentioned in his earlier testing with Bullet that he'd discovered the high poly count really didn't seem to increase simulation times, and it was actually a matter of a smoother mesh that gave less problems and faster/more accurate sims. Which is weird, because in many other instances, high poly count leads to longer sims (I'm thinking of dynamic hair) so maybe the opposite of what you'd expect.

    By the way, PhilW mentioned in one of his vids or maybe it was a forum thread, that Bullet cloth really prefers Tri instead of Quad polygons for faster and more stable cloth sims. Just throwing that out there to see if others can confirm or if it may be helpful.

  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited February 2015

    Jonstark said:
    By the way, PhilW mentioned in one of his vids or maybe it was a forum thread, that Bullet cloth really prefers tri instead of quad polygons for faster and more stable cloth sims. Just throwing that out there to see if others can confirm or if it may be helpful

    Sorry, but I disagree with that too. I believe that the Bullet sim converts everything to tris anyway. And my testing long ago showed, as I recall, it made absolutely no difference, and in fact I had a few cases where tris took a bit longer.

    So, in my experience and testing, the "tris are better" and "low poly proxies are better" beliefs generally don't seem to be true. I recall even posting my results long ago, and I don't recall anyone posting results contradicting what I posted.

    Again, I encourage anyone to do their own tests, and I'll be happy to change my position if the facts warrant. However, based on the personalities involved, I fully expect those beliefs to continue as folklore... :) :) :)

    Post edited by JoeMamma2000 on
  • JonstarkJonstark Posts: 2,738
    edited December 1969

    I think Diomede's testing seems to confirm your point about low poly not making a difference. Interesting about Bullet auto converting to tris anyway, so in that case it would make no material difference if you took an extra step and converted to tris first.

    Too late for me to play and test tonight though... But I'm very excited about the progress Diomede is demonstrating, this is looking very viable indeed. I haven't done much to help test or learn stuff yet, hopefully I'll find some time tomorrow to mess around.

  • JoeMamma2000JoeMamma2000 Posts: 2,615
    edited December 1969

    Jonstark said:
    Interesting about Bullet auto converting to tris anyway, so in that case it would make no material difference if you took an extra step and converted to tris first..

    I'm not positive about the converting to tris...I could be off on that.

    As far as the low poly vs. high poly. I've had some very low poly, nasty mesh that takes forever to solve, or even crashes. And I've had a lot of suprisingly high poly, very clean subdivision mesh, with multiple smoothing levels, that solves super fast. I recently posted an image of a bullet cloth result on a character (the blonde with the pink gown) that went very fast. So like I say, it depends more on mesh quality than resolution.

    And that's simple and easy to test and verify.

  • PhilWPhilW Posts: 5,145
    edited February 2015

    It would be very easy for someone to do a quad mesh, drape it to see how it works, and then triangulate and see if that works the same? My findings on tris originated from importing clothing from Marvelous Designer 2 which only exports tris - the latest version supports quad meshes, albeit sometimes pretty messy quads. Sorry I can't do this myself right now.

    Now that the exploding undermesh issue has been apparently resolved, I was thinking about joints. When I was able to do a test before, there were issues with the undermesh parts not connecting and therefore the cloth draping between the parts. This seemed a particular issue at the shoulders, but knees could also be affected. I was wondering if you could make capsule shaped parts that overlapped - without self-collision this may not be an issue and it would deal better with the gaps in the joints by having them overlap. Maybe someone could try it?

    Oh, by the way - great progress!!

    P.P.S. If you export and then import the same mesh, disable Auto-position and Auto-Scaling (so reverse the default check marks on these) and it should come in exactly the size and position it was exported from.

    Post edited by PhilW on
  • DiomedeDiomede Posts: 15,168
    edited February 2015

    All hail Stringtheory

    I will try to summarize what I think the answers are to some of these questions and what I think the state of best practice is. I emphasize think because I don't know for sure, and I encourage people to keep experimenting and sharing results, both good and bad. Attempts to replicate are good. Going up a blind alley or two is not bad. Disagreeing with what follows is not bad. Demonstrating that current thinking is erroneous (or correct!) is good.

    - Stringtheory's basic concept is that a softbody envelope can be created around a regular keyframed rigged figure. So Stringtheory created a softbody cloth obj with inner panels softbody attached to bones of a rigged figure (brilliant!) and an outer mesh that was the cloth. Collision with other objects for the underlying figure was turned off so that only softbody against softbody (in this case self collision for the vertex object that has both a set of inner softbody attach panels and an outer clothing mesh) needed to be calculated.

    To make it all work a soft body modifier is added to the vertex object and the inner mesh sections are then each individually “soft body attached” to the corresponding bone section of the character. This is done by adding multiple “soft body attach” modifiers and then using the “Named Vertices” button to pick the correct set of vertices (see image below). One note of caution is that multiple shading domains seem to interfere with the named vertices when it comes to selecting the vertices for attachment.


    - to me, the big discovery is that Carrara seems to calculate softbody collision with softbody much faster than softbody collision with a regular figure. So, one question was whether this was driven by the lower poly count of the envelope. I have a test of that proposition on page 6 of the thread, but it is just one test, so I encourage people to look at what I did and try different tests with different figures and different softbody. I don't think it is the lower poly count, nor does mesh quality matter much for calculation time, unless the term quality is being used in a way I don't understand. It is softbody collision with softbody that is getting calculated faster. But replication is important so please, please, please report further tests if you disagree.

    - one question that immediately follows is what is the best way to have a softbody envelope match the movement of the underlying keyframed figure? Again, Stringtheory showed how softbody attach could be applied to boned bodyparts in the first post.

    The body suite mesh is then separated at the bone joints so that each major bone section has its own separate exo-shell. For my POC I just took the V4 body suite mesh and chopped it up but I think a custom made low poly mesh would be more effective (I’ll work on one of these as time permits). Each of the different inner mesh sections needs to be given a vertex name.

    So here is where I think things stand

    General concept of cloth and undersuit process

    - the softbody undersuit and the softbody cloth do NOT need to be in the same vertex object.
    - the softbody undersuit and the softbody cloth SHOULD NOT be in the same vertex object because then you can preserve self collision for the cloth but turn off self collision for the undersuit
    - the different softbody panels corresponding to figure bones CAN be in the same vertex object with each other (different from clothing), then turn off "self collision"
    - the different softbody panels corresponding to figure bones SHOULD be in the same vertex object (different from clothing), because you can simultaneously turn off self collision but preserve collision with the cloth.
    - As long as you can turn off self collision, it is OK for the softbody attach panels to intersect when joints move if they are in a single vertex object.
    - There are still some imperfections (eg. minor poke through) in some cases because there is still no weightmapped blending of the panels during joint rotation. In the future this might be addressed with innovation in undersuit design at the joints.

    General approach to underarmor construction
    - you can create your own panels if you want - best to have in single vertex object (EDIT: try to match shape of bodypart)
    - if you are not a modeler, for a given figure, you can export and reimport (twice!) the figure mesh to get an obj with separate grouped bodyparts. You still need to name the vertices because the import process will name polygons, not vertices (I think)
    - you may want to increase scale on particular axes to give a bigger margin of error around bodyparts (X and Y for shin, for example)
    - if you use the export/import process, you probably CANNOT share the mesh. You probably CAN share the undersuit mesh if you create your own.

    General approach to clothing construction (settings for different materials, different styles, etc.)
    - not my department, said Verner Von Braun
    - see Marcus, Joe, Stringtheory, and many others
    - I would note that you can still softbody attach a portion of the cloth to the rigged figure (such as a belt)


    Hope that helps. Keep experimenting and keep reporting. Like I said, I could be wrong.

    Post edited by Diomede on
  • DiomedeDiomede Posts: 15,168
    edited December 1969

    diomede64 said:
    Yes, good point. At that point I prefer to drape using Poser. Faster, less hassle, and I'm more familiar with the tools.

    I'm curious...how do you get your Carrara scene into Poser to do the drape?

    It is pretty straightforward because Carrara's browser supports Poser's runtime structure. That means the exact same pose can be saved in Poser then applied to the figure in Carrara. Similarly, the completed drape on the pose can be saved in the props folder then loaded directly in Carrara.

    The details are here.

    http://www.daz3d.com/forums/discussion/45361/#675785

  • DiomedeDiomede Posts: 15,168
    edited February 2015

    Roygee said:
    @diomede - that is starting to look really viable :)

    Couple of questions, if I may. The imported .obj will have the same mesh density as the original - so it appears that it is not the mesh density that is the problem with collision, but "something else"? Any idea what that something else may be?

    Secondly - just a suggestion. If you export the .obj, then import as a single .obj, delete the parts not needed, name vertex groups, export as .obj and import as .obj as separate groups, would that not cut down on all that copy/pasting time?

    RE: why the difference? I'm not sure. I suspect it has something to do with how softbody is calculated but that can't be the only factor because we see so many tablecloths draped over cubes. Hopefully, someone with a better understanding of physics calculations will drop by.

    RE: second import. Excellent suggestion! Note - will still have to go through and name the vertices of each panel but a great time saver.

    Post edited by Diomede on
  • DiomedeDiomede Posts: 15,168
    edited December 1969

    BTW, speaking of Marvelous Designer....

    From a thread I started about 2-1/2 years ago, July 2012, which I started out by saying "Since Bullet is pretty much dead in the water for the forseeable future..." :) :) ....

    Here's an image of some clothing developed very quickly in Marvelous Designer, which at the time was quite new, and the guy doing it (me) had just started using it. So apparently it was idiot proof.

    The wonderful thing about it is that you design and build your clothes using a method very close to what real clothing makers use....you lay out the pieces on a flat, 2D pattern, and it automatically assembles the pieces at the seams and does an excellent, real time shrink fit. And the UV maps are wonderful, matching the patterns the way you want them too, and the way they do with real clothes.

    Stuff like the gown thing with all the ruffles shown below can be done in minutes. And if you know what you're doing (which I don't when it comes to clothing design), I'm sure you can make stuff that is incredible. And I'm sure it has been vastly improved over what they had in 2012.

    I dunno, it just seems like a nice dream for the future of Bullet....

    Sorry for the diversion into dreamland...

    Marvelous Designer looks great. I think they have an extended trial version going now if people are interested. And, I think one of the plugin creators has an obj sequencer to facilitate use with Carrara.

  • DiomedeDiomede Posts: 15,168
    edited December 1969

    Roygee said:
    Secondly - just a suggestion. If you export the .obj, then import as a single .obj, delete the parts not needed, name vertex groups, export as .obj and import as .obj as separate groups, would that not cut down on all that copy/pasting time?

    I wrote a procedure for importing the .obj as separate parts right on top of the actor/avatar/rigged V4/whatever you call it. So no need to copy and paste anything. All the parts are named like the rigged character.

    Or maybe I'm not understanding his procedure...

    Yeah, I missed a checkbox. RE: the names. The softbody attach vertex selector needs the vertices named. It won't allow you to select by polygons. Aaargh. Something to ask for.

  • DiomedeDiomede Posts: 15,168
    edited December 1969

    Jonstark said:
    Roygee said:

    Couple of questions, if I may. The imported .obj will have the same mesh density as the original - so it appears that it is not the mesh density that is the problem with collision, but "something else"? Any idea what that something else may be?

    Joe mentioned in his earlier testing with Bullet that he'd discovered the high poly count really didn't seem to increase simulation times, and it was actually a matter of a smoother mesh that gave less problems and faster/more accurate sims. Which is weird, because in many other instances, high poly count leads to longer sims (I'm thinking of dynamic hair) so maybe the opposite of what you'd expect.

    By the way, PhilW mentioned in one of his vids or maybe it was a forum thread, that Bullet cloth really prefers Tri instead of Quad polygons for faster and more stable cloth sims. Just throwing that out there to see if others can confirm or if it may be helpful.

    I agree. I think the primary time savings is not from polygon count. It is from softbody against softbody instead of softbody against rigged figure. I have a test of this proposition using A3 going to a sitting position on page 6, but welcome other tests with other figures and other cloth construction (including the mesh quality issue).

Sign In or Register to comment.