Adding to Cart…
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.You currently have no notifications.
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
Thanks Jonstark,
I hope I haven't diverted the main progress of this thread down a blind alley by the bones musings or by showing some drapes on static poses.
However I could end with a couple of observations on the gifs
1 the white garment was essentially a cylinder with the lower part flared out into a disc. It was flattened when placed on the figure because the figure's torso is wider than it is deep. That meant that the 'disc' became almond shaped rather than round. When draped, the sides are longer than the front and back as a result.
2 the same garment has some crinkles mid-way down from the waist. That was where I triangulated some rings of the disc. The rest is quads.
3 where the white clothing starts to flare out from being a cylinder is the area where 'pleats' start to appear during the drape.
4 in other examples I tried, some of the rings of quads can be longer than the other quads - they will also make a difference to the final drape - such quads will stretch a bit more than the others.
5 on the long pink garment simulation I moved the character's arms down after the simulation. The cloth was just an animated object at that point.
Oops, Diomede, I didn't see your comments - you're very kind and I'm glad you like the pics.
The one you liked best uses, of course, a few soft-body-attached vertices (I wish I had a quicker term for this!)
As such, they could be moved back and forward if attached to something. The nice thing is that the arms don't need to be animated until the simulation is over so they can be kept out of the way of colliding. The simulation leaves an animated object which can be exported as a still .obj at any frame.
Remember that gravity can be positive or negative and can be set in X,Y,Z to resemble a wind-force (or even a partial shrink-wrap?).
I wonder if separate panels of a cloth could be placed against a figure one by one and draped by horizontal gravity? Then exported as solid objects?
Once all the panels were complete, assemble them in the VM. (Very) poor man's Marvelous Designer?
Oh no! Time for my medication.
Actually I need to log off now. Thanks to all.
diomede asked me to do some further testing in Blender on the Poser figures.
Started with James - no problem rigging and automatic weighting. I haven't made him any clothing, but am sure that will work well, because of what follows.
I then went on to Aiko3 and 4. Again, no problem rigging, automatic weighting, shrink-wrap and clothing sim, using the same dress as previously used for V4. Now this totally puzzled me, because Aiko 4 is a clone of V4.
So, I made a new export of V4 and started fresh. Same problem as before with the rigging - automatic weighting cannot identify vertex groups to weight-paint. No problem this time with shrink-wrap or a static cloth sim. I'm guessing that if I do manual weight-painting to make animation possible, a good animated cloth sim would be possible.
I'll try that when I have time - it is not all that difficult in Blender; I just don't have much free time at the moment.
One thing I did differently when exporting the .obj's from DS was to weld the groups. This probably made the difference and here's something to ponder in the think tank.
There was a question earlier about the bones in Daz/Poser figures and the difference between them and Carrara rigging.
All armatures, skeletons, call them what you will, work the same way in all applications. Groups of vertices are allocated to bones. The difference between the rest of the world and Daz/Poser rigging is that in all other applications the rigging is drawn in the figure and influences either automatically calculated by a sphere of influence based on distance from the bone, or painted on, or a combination. With Daz/Poser figures, the vertex groups are assigned beforehand and the bones generated to match.
Now this may or may not be significant in the problems experienced by doing an animated cloth sim in Carrara and getting automatic weighting to work in Blender on V4, M4 and Genesis. Before the DS-Hexagon bridge, we had the devil of a time doing full-body morphs, because each body part was a discrete mesh. To make a .cr2 figure, we have to physically cut up a figure into body parts to attach bones, or allocate vertex groups using other software.
Nothing really changed with rigging for Genesis, except to introduce Tri-Ax, which allows different weighting to be applied to each axis. I personally feel that Daz missed a great opportunity when developing Genesis to change the rigging system to what the rest of the world does.
Anyhow - the point to ponder is whether the clothing sim is getting confused by all these bits of body parts and when you attach the underarmour, you are actually protecting the clothing sim from this confusion by giving it a single object to collide with.
Which is why I'd be interested to see whether a single-mesh figure, rigged in Carrara, would give a simpler sim. I can provide a rigged figure, (Janet - my crash-test dummy - with a dress, if someone wants to test. No good me testing, because I'm on 8.1.
It would also be interesting to see a test done on a Collada export from DS - because it works so well in Blender, it may just work in Carrara.
A couple of points which just occurred to me, which you guys may want to think about.
Firstly, in Blender, you can set the sim to start at any point, or even before the start of the timeline, to give the cloth time to settle before the animation starts. Maybe freeze your figure for a couple of seconds before starting the animation.
Secondly, if I lay my home-made and Carrara-rigged figure supine, I can drop a sheet over her and it drapes nicely. Put a dress on and it goes moggy. The only real difference is that the dress encloses her and the sheet doesn't.
Testing a theory
The first segment shows the hammer without rigging - movement is by parenting and rotation. The movement is pretty much what you could expect, using default settings.
The second segment has the hammer with rigging, set to collide with the blanket. Not too bad, but not as "natural" as the first.
The third has the hammer with rigging, set not to collide - it really goes wild and tries to get away from the rigging, but is held back by the soft body attachment to the rod.
The third has the head of the hammer covered with a soft body attachment. This is protecting the blanket from the rigging, and is pretty much the same as the second segment.
Ipso facto - the rigging is the culprit in this whole drama!
:)
http://youtu.be/oonoM9saPC8
Roy, great experiment! Thanks so much for your latest posts. Very informative.
I just woke up so I'm not sure my brain has fully processed it all, but I think I now have a couple of related tests that I want to try with a couple of my own rigged figures. More on that when I get a chance, but if it is the rigging that is the key then that explains why merely creating an envelope with a simple attach-to-skeleton in the animation tab did not work in my early tests.
The following is just about convenience, not whether something works or not. For those interested in using the figure's own mesh, or a low-poly development rig that comes with the figure as the base for the undersuit, it can be tedious to assign the vertex groups. The vertex modeler allows for naming by polymesh, polygon, edge, or vertex. The mesh for V4 and similar figures has its bodyparts named by polygon, not by vertex. The softbody attach menu assigns groups of vertices, not polygons, to the skeleton tree, and so quite reasonably allows for selection by vertex group, not polygon group. Is there a way to simplify naming vertices groups for a mesh that has a set of polygon groups?
Roy, fascinating stuff. Ironically I was up late trying another approach of putting together a smoother and more form-fitting underarmor on my V4, but wasn't online looking at the forum, so only catching up to all this new info this morning.
I'd be happy to volunteer to test out your carrara-rigged/created character in 8.5 to see what the cloth sims are like. I'm on a break from work at the moment, so I wouldn't be able to do anything til much later today or tomorrow, but nevertheless I'm very interested to test the theory out, everything you've mentioned makes me thing the logic behind your thinking is sound though.
I'm still trying to find an effect method of underarmor that will work well and still stay 'skin tight' in any posed position. Having trouble deciding what to do about the glutes/buttocks. There are advantages/disadvantages to attaching the proxy glutes to either the thigh or the hip, not sure what the most ideal solution would be. For most walking and standing I think thigh would be best, but if I have my character bend way over at the waist or squat down, not sure but think maybe hip is better... I'm sure there must be a good solution, just not seeing what it is clearly yet.
Not sure if it's of any interest, but since all info is useful in some ways, I'll post my current underarmor rig that I'm fooling around with:
After proving I am not very good at all at modeling with primitive cylinders, even though that method does work as per my prior tests, every time I went back to try to make the mesh smoother and better shaped I seemed to only mess it up more :)
So I went back to the method Diomede mentioned of using a mesh shaped like V4 and in this case imported the low poly LOD V4 (the 4k version) then made it about the same size as my character, went into the modeling room and deleted a bunch of polys in places that made sense to my mind logically (but might not be right!) so there would be different bodypart sections that could move independently of each other and be soft attached to different bones. Ironically the mesh already has named polygon sections, not sure how to get rid of that, but it doesn't matter since the soft attach works with named vertexes, and it was easy enough to select each section and name it's vertexes for the body part it represented.
I'm probably only going to use the abdomen and below mostly, so that's the part I concentrated on most, it was relatively easy with symmetry and soft select to 'pull' the various parts in to place so that it fit better to my specific character (she wasn't really all that far off from the default V4 shape really). This method of using a V4 mesh obj is much much easier to get a nice smoothly conforming 'skin tight' underarmor I think. My primitives and cylinder method will work, but this way is much easier IMO, so for now I think I'm abandoning the primitive cylinder method :)
Some thoughts:
Collars/Neck/Chest: I left the chest, neck, and collars all as one piece to attach to the chest. I could have broken it up further but didn't really see much point as the collars won't move much anyway and while the neck might move around rotating and bending a bit, I don't think that would matter much except for the most demanding turtleneck dress. But I might split those up on down the line.
Elbows: I haven't done the elbows yet, but I'm imagining something like what I've done with the knees might be just as good a solution, in other words leave a space between the forearm and the shoulder pieces and insert a primitive sphere, shape it a little and have it attach to either the shoulder or the forearm bone (I think either could work equally well, but leaning towards the shoulder).
Knees: Left a gap between the shin and thigh pieces and stuck in a sphere, shaped it a little to fit better, and have it sticking out enough that I think it should give a nice knee effect and also more importantly the cloth shouldn't poke through the cracks and crevices.
Glutes: Still struggling with finding an ideal solution for this area, as the butt can be influenced by the thighs or the hip bones. I've got them as part of the thigh bodypart currently, not sure that's the best solution but I do think for walking and standing thighbone movement is more influential.
Thighs: You'll see that it encompasses both the thigh and also the side of the hip and the glute above. I'm unsure if this would be the best approach, as I'm worried that the side of the hip might need to follow the hip bone more than the thighs...
Looking good, Jonstark. This is just a suggestion for convenience. It won't actually help performance.
Even if you plan to make your own custom shaped capsules, you might still want to export and re-import the figure mesh with the option of creating an object per group. The reason is that you can use modeling in the assembly room and check and uncheck the visibility of the bodyparts that surround the panel you are working on. That way you can get a very good approximation of where the mesh should extend. For some bodyparts that extend lower in front than back, or vice versa, it can be a help. For example, if you are making a panel for the right forearm, uncheck the visibility of the imported right shoulder and right hand, and then make the capsule around the right forearm. You can then select that whole capsule and name the vertices "Forearm R" or something. Exit the undersuit modeling, then make the right shoulder visible but the forearm invisible and the right collar invisible. Select the undersuit mesh and enter vertex modeling in the assembly room and create a new distinct capsule for the right shoulder. Name the vertices. And repeat. I think you can get rounded capsules that way if you want and still have them generally correspond to the figure bodypart shapes. Like I said, more of a convenience thing than anything else.
It continues to be interesting and it's impressive to see how determined everyone working on this is.
As mentioned, the figure I mostly used in experiments was exported from MakeHuman 1.6.
I took the mesh into Hexagon and unified it into one-piece.
In Carrara's VM I found it had a whole shell of vertices around it not joined by edges. ??? Perhaps the developers had clothing in mind? It caused problems for me because I posted a whole video on draping where I focussed a bit on how the mesh crinkled badly here and there. Only later did I discover that there was a second mesh of vertices which must have affected the simulations.
Strangely, I recall that I only had to select a couple of these extra vertices and, by deleting them, the whole lot disappeared.
Anyway, I put a very simple set of Carrara bones into the mesh - missing out fingers and toes - and tried my hand at weight-painting.
Now, the nice thing about such a figure is that you can detach the skeleton. See where this is going? With the skeleton detached you could delete, say, the whole torso and re-attach what's left onto the skeleton. If the torso is going to be hidden by clothing, it doesn't need to be there. But it could be chopped out, decimated and put back as a lower-poly collision object. In Carrara's VM there is an option to decimate and keep the shape.
If all the parts covered by clothing, such as the thighs, were one group, self collision could be switched off and they could be scaled a little to overlap.
I can add some weird images of the concept. I started here with a skeleton and united it with a mesh which consisted of only a group of blue cylinders. These could represent the nicely-modelled parts of a figure. (one is twisted because I changed joint orientation afterwards.)
The images also show, I'm sorry to say, the problem of a collision sphere leaping upwards when I rotated the whole figure from the hip.
You can see how it rotated from the bone - as though it's pivot-point was there too.
@ Roygee - very interesting. I like the hammer hitting the cloth just as an animation regardless of the purpose.
@ Jonstark, thanks for the helpful comments yesterday you made when I was talking about bones. I agreed with all you said but forgot to acknowledge that. I don't think you have any apologies to make for your modelling skills. It all looks excellent to me.
This might sound negative to you or anyone else reading but I hope you don't regret using precious time off on zillions of simulations. I did that and ended by giving up eventually. It always seems so close but then eludes us again. Meanwhile Carrara is still able to do all the other things you love about it.
I could liken the situation to using a mail-merge in a word-processor. You pull down a menu for that task and follow steps 1-5 and the merge takes place. Poser's cloth room is rather like that. It's presented to us to use.
If doing the same thing in Carrara needs a million trials and adjustments then it really isn't something that a client, say, could be told is do-able in a decent time-frame.
YET, being optimistic again, those in the thread trying out everything possible are making progress for sure and may well get to the point where we all can have a good idea of the step-by-step method. Carrara does have impressive Bullet tools but the settings are still in the wild - maybe they can be harnessed and tamed.
One thing, when I first tried Poser's cloth room, I found reading the help files like reading a children's book after all I had tried in Carrara. So what everyone can learn is worthwhile, I'd say. The knowledge, I guess, could be carried over to other programs.
@jonstark - I've sent you a PM for the location of Janet:)
@ Marcus Severus - just a bit of fun, thanks :)
Just struck me that this could all be wound up in a flash if a developer would come in here and let us know just where they stand on Bullett at this time, what the issues are and what work-arounds could work.
Anyone know a friendly dev who can spare a few minutes?
No-one wanting to try a Collada import from DS?
Yeah, I think you're right. Based on the fact that cloth collision with moving objects was working fine in earlier betas, I wouldn't be surprised if there was minimal work involved to restore it. Certainly a lot less than all here have been spending to search for a workaround... :) :)
Roy, I got your pm but I'm having trouble getting the download, the site keeps trying to force me to download softwares I've never heard of to get to the file. At first it was giving an error my flash needed to be upgraded and trying to upgrade my flash. I wasn't so sure I trusted what I was seeing pop up, so I closed down, went to the actual Flash site, and from there upgraded to the latest version, but the fileshare site still seems to want me to upgrade my flash with whatever their version is, also wants me to download some other smile type software.
I'm just not sure how to proceed to download the file. I don't really trust the site, but then I've never done filesharing before. But I do trust you, so if this is the correct way and all on the up and up, let me know and I'll proceed and download the different softwares the site wants me to download.
@jonstark - I have replied to your PM, hopefully showing what to do.
It has been a while since I use 4share and I see they have added a lot of "crap download" traps - don't download any of that rubbish - only the file I showed.
i'll close my account there - anyone know of a decent free file sharing host that doesn't try to trap users into downloading a bunch of rubbish?
Seems like now even the "good" filesharing sites have a lot of traps. And the common attitude is "oh, you should pay attention before you click a button". I downloaded a file from one of the well known sites, and was in a bit of a rush and didn't pay close enough attention, and ended up spending about an hour searching and destroying all of the crap it installed.
I especially hate when it infects your browser and changes your home page and search engine and some of them keep you from changing it back. What a pain.
I've always thought there should be a reverse process, where if someone infects your machine, you should have the capability to send a self destruct code to all of their servers and their entire building would go up in flames. :) :) :) That would be fair don't you think? :) :)
You could try using the Public area of Dropbox.
Just to share something weird and unexpected I ran into that may be of interest (under the theory that all info on this subject can be useful somehow).
Last night after I bought Aniblock Importer for Carrara and installed it, I had to close Carrara and reopen to enable the Aniblock importer. Before I did that I saved my scene with the softbody underarmor that I built for my V4 character (pictured on the earlier page).
When I went to run a clothing sim after importing an aniblock, I got 'an error has occurred'. At first I was dismayed thinking maybe there is something inherently about this method that doesn't like aniblocks, and that made zero sense to me since they are just glorified pose files, and nothing so far has dismayed the cloth sim from running with whatever poses I chose, but then I looked closer into the matter, deleting and restarting the scene from the original save point and not using an aniblock, and discovered the problem was that after I saved the scene file and loaded it back up again, it did not save my soft body attaches in the right place, and each section had random vertexes attached to each bone from all over the body parts. Meaning every softbody part of the underarmor was attached to each bone by random vertexes. So it looks like maybe the underarmor doesn't save the soft body attaches the way they were set? Not sure... I think I should have dragged/dropped V4 and the underarmor into the objects tab to get a cleaner save, yikes...
It'll take a bit of time to fix it so the vertexes are attached to the right bones, not a big deal but still threw me for a loop so thought I would mention it.
Roy, I'll stand by to see if there's another way to get the carrara file from you without getting forced towards downloading some weird files :) I haven't had much time to test anyway and likely won't til the weekend, but I am looking forward to testing as hopefully it will uncover another piece of the puzzle.
OK - I've posted it to dropbox and sent you the link. I used to use dropbox until that scare with the Heartbleed virus. Problem was I never could find out how to share without knowing the recipient's e-mail address. I see that it has now been made easy.
It's a nice clean site, without any funny stuff on it, as 4shared used to be :)
I am so glad that you posted this. I think it helps me understand what went wrong with a full Poser 7 figure with multiple panels that I was trying in this post http://www.daz3d.com/forums/discussion/49954/P285/#763053
As you might expect, I did the test for the thigh first, then when I had soft-body attached all of the undersuit panels for the full figure, I grouped the P7 figure and the undersuit softbody attach together and saved the group to my browser, started a new scene and loaded the group and tried the physics sim. I got the same error you did but just thought I had made an error applying the other panels. It may be that something strange happens to softbody attach in the save and load process. Maybe, maybe not. To me, it is still worth exploring.
So, yes, if folks are still experimenting, please keep posting both successes and failures.
The dropbox worked great Roy, I've got her now. Hopefully later ttoday I can get a chance to play :)
JoeMamma2000 I do relate to your feelings regarding web-sites that make changes to our computers. Mistrust has become the order of the day for me.
Particularly I hate being required to give a mobile number and email address before I can complete a transaction on-line, such as buying a train ticket. I look upon my email address as something I pay for - its for me and those I choose to give it to. It shouldn't be public domain like my street address - especially since it will be sold on. As for mobiles, I hate the things and I never have the number to hand for the pay-as-you-go mobile I use about twice a year. The only thing some web-sites don't ask for is colour of underwear being worn, it seems to me.
RANT OVER
I found it interesting that collision worked properly in the Betas. I hadn't bought Carrara at that time. Were the sliders, tick-boxes, etc. the same in the Betas as we have now in 8.5? If so, something just got broken?
To Diomede and Jonstark. I noticed that if a mesh with some of its vertices painted as soft-body-attach is changed (eg some faces are deleted, then the painting has to be re-done).
Roy, I just realized you outfitted her with a cloth dress already, testing this might be the easiest thing I do :) You've got it all set up and ready to go, I might even be able to do this while I'm at work today, depending on how busy things are (I work from home, with my work laptop to the left and my play laptop to the right, so when it's slow I can tinker in Carrara and check forums. Not a bad deal, all things considered).
@jonstark - looking forward to seeing whether native Carrara rigging makes a difference :)
As a test, I brought in James as a .dae export from DS and made a quick skirt - it certainly behaves much better than anything else I've tried in Carrara - no flying off into space, even without soft body attachments and it seems to want to move the cloth with the limbs. Just need to get more refined settings.
I noticed you guys keep referring to "margin" - 8.1 doesn't have this setting - is this maybe something like collision distance?
Regarding sites that try to trick you into downloading crap - yes, we must be ever vigilant. Something which totally amazes me is that there appear to be marketers who think they will get business by constantly annoying potential customers. Really weird philosophy!
In my country we have seriously good consumer protection laws - every unsolicited marketing e-mail or text message has to be accompanied by an option to opt out. Virtually every time I open e-mail, I spend more time deleting and opting out than reading! Can't do much about those from foreign countries, though - just delete, delete, delete.
My quick run of a test dummy rigged natively in Carrara was not very successful.
- I modeled a low-poly human (barely a face!)
- I modeled a simple skirt
- I named some vertices "belt" at the top of the skirt
Scene settings
- Physics simulation accuracy 375
- Geometric fidelity 97
Testdummy
- collisions on
- other settings default
- Simple bone rig and skeleton attached
Skirt
- softbody settings
- stiffness 80, bending 1.8, self collision on, margin 30
- softbody attached belt to hip of rig
It took forever compared to the softbody envelope
The leg went through the skirt
Diomede, looking at your scene tab, I don't see any collision mesh. As I was suggesting yesterday you could try:
1 detach the skeleton
2 chop out the thigh geometry (for example). Just leave a gap. It will be covered by the clothing.
3 Re-attach the skeleton.
4 take the thigh mesh you chopped out and turn it into a soft-body collision mesh attached to the thigh bone.
Yesterday I showed a little gif of the idea (using spheres). It did work but was poor quality.
BTW where I showed the sphere flipped upwards: the bone involved was one I created using 'duplicate with symmetry'. I also re-oriented that joint later. I've heard of Euler angles (?) also. Maybe some of those played a part.
Much earlier in the day I decided to do a quick sim on the figure Roy set up, which is created and rigged in Carrara with a dress object on it. For the first test I simply made the dress cloth by putting a soft body modifier on it with very low bending and low stiffness and a margin checked at 100% (I do believe margin functions exactly collision distance btw), no other changes.
Did not put any softbody underarmor on the figure, also left it as it was without any softbody on the figure either, wanted to see what would happen and since it was only a 2 second animation thought that no matter how dire it would be quick.
Wow was I wrong. Everything slowed to a crawl. In fact I thought at first Carrara had crashed until several minutes later it showed the status bar, at .01%.
The simulation is still running, low these many hours later (it's about 88% completed now). There has been pokethrough on the cloth, but it isn't flying off into space, still it's not working (and taking a long time to not work, lol). I have hit the esc button to try to halt the sim, but it ignored it. I could have crashed out of Carrara to stop things, but I had another project open in a different window that I didn't save just prior, plus I'm still at work so no big hurry anyhow, so I"m just waiting it out.
But in general sounds like I'm getting very similar results to what Diomede got in the experiment he posted just above.
What Phil said. Dropbox's public folder. For me at any rate, if somebody sends me a link to a file in the dropbox public folder, all I can do is download the linked file. I can't see anything else in the folder, so in that respect, it seems pretty safe. It doesn't like spaces in the name, so when you add a file to it, make sure the name doesn't have a space.