Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
The export/import and use of the original mesh can work fine, as Diomede demonstrated, but it will need to be altered somewhat, it won't be enough to simply have the pieces separate from each other. Because although the mesh names of the various group are named the same as the bones, they don't actually conform to the mesh parts that are moved by those bones. For example, the named-mesh of the collars extends well down the shoulder, nearly haflway down the shoulder in fact. Then the shoulder mesh actually lies on the lower half of the shoulder, the elbow, and part of the forearm. Just one example, and you've already highlighted what the hip mesh looks like, but in reality I think the actual part that needs to be able to move with the hip 'bone' movements is really just the part that we would think of as the waistline.
Seeing this, and being a little daunted at the task of trying to decide which polys to delete (so that various parts didn't 'rub' into each other when the character was in motion, which I found can cause mesh deformation even when the 'collide with self' is unchecked in the softbody, I decided to try out a less precise for the surface of the skin and to my mind cruder approach.
I used vertex room simple cylinders, each one is a contained object btw, with rounded tops and bottoms, and placed each cylinder along the path of an actual bone (not the named section of the mesh but the actual influenced part of the bone movement), and used a little dynamic extrusion/extract along/etc to make each cylinder rounded at the top and bottom rather than a flat circle. Then I just matched it up to my V4 in the T pose in the assembly room, used 'model in assembly room' and moved with soft selection here and there until it covered the relevant body part over each particular bone.
I was worried about the joints, decided to make the knees an extra 'ball' at the bottom of the thigh cylinder, did something very similar for the glutes up above. I would theorize that the knees could also just as easily be attached to the shin bone, but either way I believe they are very important to have a sort of 'ball' joint (at least protruded and that will move in a curved way when the leg bends). Obviously my approach was very rough (kind of incredibly ugly, and real modelers are turning away in disgust at the sight :) ) but it's invisible in the actual render anyway, and the cloth seemed to work well with it. I'm going to take a crack at making a more refined approach, but I'm incredibly jazzed at how well this seemed to work at a first rough try. Almost seems like it's all down to figuring out the settings for the different cloth types, because the rest of it works (and the cloth sim was really, really fast too, as an unexpected bonus). I'll see if it was blind luck or if I can replicate it and improve with an even cleaner mesh for the soft-underarmor.
And here's the under armor alone, with the main character visibility turned off...
I scrunched (scaled) each part a little in the Z axis to minimize interaction.
With all that said, it doesn't work real well. Stuff explodes, like before. I may have missed a key setting...
Marcus Severus, I have watched many of your videos in trying to familiarize myself, but I somehow never saw that one before! Very instructive, you may have been the first to glean this method, and I loved the method of making V4 able to sit on her dynamic skirt :)
Another field I completely forgot about, that you demonstrated is very important in all this is the effects tab of the object and the physical properties. In my experiments with dynamic hair I forgot about this field too (except that I turned off collisions for my high poly objects in the scenes I was running hair sims on), and Dart brought it to my attention, this field, particularly bounce and friction, looks to be very important to determining the way the cloth will behave and react as it collides with the softbody object too.
Thanks JS...it will be interesting to see what, if anything, is the magic key here...
It will be unfortunate if it requires that we build proxy objects for under armor...a lot more work, especially if you start morphing characters, and if you want close-fitting cloth, which you usually want.
BTW, are you guys turning OFF collide with other objects on the main character?
Yes, I discussed this a long time ago in my tutorial on Bullet ridgid and cloth sims. The more bounce, the more collisions must be calculated, and, especially with rigid body sims, the higher likelihood for stuff going unstable and falling thru or off into space.
EDIT: Here's a link from a couple years ago:
http://www.daz3d.com/forums/discussion/23632/
Thanks Joe, right into my bookmarks :) I believe refining settings will be the next frontier on this.
There must have been some improvements made for 8.5 - I'm on 8.1 and still have the old "shoot off into space thing" that we had when Bullett was first introduced. You guys can at least get a static drape on a Daz figure. I've tried the underarmour, with all the recommended settings and that helps to a very small degree - it doessn't shoot off into space, but hikes up around the waist and sits there shivering and shaking:)
Even using my own, unrigged model doesn't work very well.
Obviously wants to shoot off into space, but is held back by the soft body attachment.
Any further testing by me would be invalid for what you guys are doing, so I thought I'd give it a go in Blender, as they use the same Bullett physics and I may be able to help uncover the secret.
Discovered some quite startling things.
Firstly, Bullett simply does not enjoy Daz figures. I started using v4 and G2F .obj exports and found that the rigging wouldn't accept the figures. Tried simply draping on an unrigged .obj; takes forever. It did work for me in the last challenge on an unrigged M4 .obj, but that was a cloak, so the figure was not enclosed in cloth.
When I tried to shrinkwrap on a Daz figure .obj, it simply cancelled the operation!
Experts in the Blenerartists forum advised me not to use Daz .obj's and I found that, sure enough, it works fine on my own and other human forms randomly found on the internet.
I was advised to use either Collada or FBX exports from DS. Collada worked fine, after deleting the weird skeleton and making my own rig.
The cloth sim ran at almost real time and the results are really good, using default settings for silk cloth. Please bear i'n mind that I'm still very much a novice in Blender and have concentrated on modelling; still unraveling the mysteries of animation:)
Next I'll try a longer, more complex anim and do a hair test.
May be worthwhile for you guys to try importing and clothing a Collada export from DS.
Have any of you tried doing a cloth sim on a non-Daz figure? Would be interesting to see your results:)
Anyway, here's G2F doing her thing in Blender :)
https://www.youtube.com/watch?v=gTWluIxcRB0
I wonder if that's why I'm getting such fast cloth sims, Roy, since I've got the daz figure unchecked for any collisions and I've got it instead colliding with primitives... I think Joe has already found good evidence that the poly count doesn't matter, but maybe there's something inherent to the daz stock characters that slows things down? It's a weird effect, nevertheless. I was shocked by how fast the cloth sim ran for me, like you said, nearly real time.
Nice looking drape/dress in that vid there :)
Couple of suggestions.
- if your under armor explodes, make sure that (a) self collision is unchecked, and (b) the settings for margin are checked and are in the 30 range. Also check the settings of your cloth article for margin. See earlier posts for sample settings. I no longer experience explosions.
- turn off collide with other objects for the base figure (V4 or whatever). Because collide with other objects for the base figure is disabled, one would think that the base figure would not matter, but it could if the base figure affects soft body attach. So far, I have successfully used the technique for V4, Aiko3, Genesis2 female default, Genesis2 female customized, M4, and David 3. Worth checking for others, and I am halfway through a test of the Poser 7 figures. My current working proposition is that the base figure does not matter in Carrara because it's collisions are disabled.
- I have had success making my own under armor suits and using the base figure exported and reimported. For me so far, the key has been getting the under suit in a single vertex object with multiple panels and self collision unchecked. So I don't think it matters if the under suit is a capsule or not, except in its shape's ability to fully prevent poke through, which also applies to under suits of other origin.
EDIT - The following thread has the settings are the ones I have had the most success with for multiple figures
http://www.daz3d.com/forums/discussion/49954/P255/#759237
Interesting...I turned off collide on the base figure, but when I did that the softbody-attached under armor pieces just drifted off into space when the simulation was run...weird. Like the base character needs collide with other objects turned on for attached pieces to stay attached? Strange...
Interesting...I turned off collide on the base figure, but when I did that the softbody-attached under armor pieces just drifted off into space when the simulation was run...weird. Like the base character needs collide with other objects turned on for attached pieces to stay attached? Strange...
Yes, this is interesting. Glad you reported it because this has worked for me with multiple figures. Hmmm, how to diagnose the source of difference?
- I am selecting the "model" level of the base figure group (V4,...), then turning off collide with other objects in the effects tab
- and during softbody attach, the undersuit's panels are attached to the bone structure
In diagnosing, just brainstorming, some questions to ask
- could the order matter? In other words does it matter if unchecking collisions is done before or after the soft body attaches?
- do you have a general softbody for the undersuit as a whole in addition to the softbody attaches?
- could you be softbody attaching to something I am not?
- are we both using the same C8.5 version (I think I have the latest 2, meaning the public version and the beta, but I should check)?
will try to think of more potential differences
Whoa that's weird Joe, why aren't the soft body attaches for the various parts to the various base figures bones keeping them from drifting? For me the soft body attach for the various parts for the various bones is what keeps it on the mesh, not the colllisions.
Diomede, did you say you are turning off collisions on the cloth object and the soft underarmor too? I hadn't thought to do that, I only have the base figure with no collisions. Very interesting if you can have them unchecked on the cloth and soft underarmor as well.
And Diomede, a huge 'thank you' for the way you've pushed this forward, I am now able to run cloth sims (fast!) that seem to work pretty great :)
Wow, so many collisions, attaches, and cloth objects to keep track of! Here is what I do.
- for base figure, I uncheck collide with other objects in the effects tab at the model level
- for underarmor, collide with other objects remains checked, but self-collision is turned off, margin is checked an at 30
- for article of clothing (dress, skirt, etc), collide with other objects remains checked, self collision is optional but I generally turn it on.
to avoid explosions
- base figure - turn off collide with other objects
- undersuit - uncheck self collision, and make sure magrin is checked and set in the 30 range - collide with other objects is enabled
- cloth - collide with other objects is on, margin is checked and around 70
Right back at you for your work on the cloth, and for the hair sims. Hopefully, we can identify some of these hiccups that Joe and others discover, diagnose them, and then establish a well defined set of best practices.
Thank you again to everyone who is reporting, and I think it is important to continue to report both successes and failures in different circumstances. More information is better. Replication is important. After my next tests, I will try to replicate your floating softbodies, Joe, but that hasn't happened to me so far. Any additional info for your settings that I am failing to ask will help me try to diagnose.
And Roy, I saw your post about Daz figures compared to others. I am close to completing some tests with Poser base figures. Let me know if P7 figures still count as a "Daz type" figures or not in terms of your Blender outcomes, and if so, suggest another type of figure for me to test.
RE: Poser 7 figure - I'll post screenshots of whatever I find, successful or unsuccessful.
RE: does the figure make a difference? Carrara does hiccup when I turn off collisions for Genesis2, but if I wait it out, Carrara returns to normal, and these hiccups do not happen for V4 or M4 or others so far.
Hi everyone,
When I posted a couple of days ago after being without Internet, or computer even, for several weeks, I had only skimmed over developments. Now having more slowly read back a few pages I have a much better idea of the phenomenal progress that has been made. Great work!
Also I jumped in with my little wooden man with attached spheres - that was sparked by seeing Jonstark's building capsules to fit around the figure's legs. With the benefit of catching up on earlier posts I can see that no-one needs to go down the weird path I was experimenting with at that point.
Jonstark - your enthusiasm is always an inspiration to me so I'm sorry to have jumped in without following on from what you'd written by congratulating you on the results. (In catching up with the forum threads I took a quick look at your first dynamic hair video - it's fantastically well done and it's a must for me to learn from that and the others. Many thanks for that.)
Sorry, Joemamma if my proxy object idea interrupted the thread and caused CORNfusion! The idea was literally the last desperate one I had last year before giving up on animating soft cloth. But the little wooden guy did enjoy his walk around town with a cone bobbing above his head!
@diomede - the summary of your findings from a few pages back is a really useful guide to all of this. Turning off self-collision of the under-armour, spheres, capsules or whatever AND making them ONE-piece is a brilliant workaround to them colliding with each other at knees, etc.
Before this thread began, I would have seen animated soft cloth in Carrara as something almost, but not quite, achievable. Now it's looking more and more that a unified way of doing things will be arrived at and good, general settings will be established. Great to see!
Oh, just had a thought, because there was one sim I ran where the soft underarmor parts went strange and fell off the figure like Joe described. Reason turned out to be because I had first soft attached all the parts, then after they were attached did some modeling to make them conform better, even adding vertexes etc, and then when I rechecked the vertices for the different parts weren't uniformly soft attached to the right bones anymore. Simply had to correct that and make sure all vertices of the various 'parts' were soft attached to the correct bones and all went back to normal.
Have no idea if this could be what you are running into Joe, but thought I should mention it.
Great work by everyone concerned. Jonstark, I hope that helps discover what is going on with the flying panels
I have good news /bad news to report regarding experiments on the Poser 7 figures.
Good news. - I did a test of just the underarmor softbody attach. I made a simple panel for the right thigh and tested whether it could be made to attach to the bone of the P7female right thigh and follow its movement with the P7 female collisions turned off. That was successful in the sense that the panel did move with the thigh, and so I expect the method can be used with the Poser figures.
Bad news #1 - The calculation time for this simple movement of one limb with one panel slowed considerably. Not sure why.
Bad news #2 - when I made a full underarmor suit for the same figure, I got an unhelpful error message, and the simulation did not complete. There are a lot of potential explanations having to do with mistakes in panel construction or use of soft body attach that could be my fault. Having said that, given the slower calculations of even the simple example, and Roy's reports of different outcomes for Daz figures in Blender, we should be on the lookout for potential issues related to the base figure. Might be worth brainstorming on different ways that a mesh might be composed, rigged, etc., and how that might be interpreted in Carara in a way that might affect softbody attach.
Anyway, I will post a screenshot of the good news - the thigh panel moving with the thigh for a non-Daz figure.
EDIT: Joe and others have presented good reason to believe that the mesh density isn't key, but just because it was easy to check, the Poser 7 female has about 40% more polygons than Victora 4.2 (about 97 thousand to about 68 thousand respectively).
Here is a link to a video I posted last March showing a sphere attached to one of the figures which came with Poser 9
http://youtu.be/HgW1UXLyBBc
Great stuff. That video reinforces the general applicability of the method. I will assume that I Made a basic mistake with the full bodysuit for P7, like inadvertently attach the same vertices to different bones or something similar.
Sorry, Diomede, my last post wasn't in any way constructive other than showing I didn't experience errors or slow-downs in that example.
I was in the middle of doing other things as well as posting!
But, for what it may be worth, since you are keen on any info for brainstorming - I did have a time when I pondered over the whole bone thing.
You see, I was mostly using a MakeHuman 1.6 mesh which I had unified into a single object and rigged using Carrara's bone tool.
At one point I tried to figure out what was taking place in attaching to a bone. I imagine the collision object (sphere, etc.) would be attached by its hot-point. But would it? Or would it be an average of the soft-body-attached vertices? If all the vertices were attached would that be the same as the hot-point?
The same question could be asked of the bone. A Carrara bone (which I was using) is placed in a mesh and is represented by a little diamond-like shape. When it's the parent of another bone there is a long set of lines representing that.
When I was attaching a sphere to a (Carrara) bone, I imagined that the long set of lines mid-way between two bones was what I was attaching to because that looked like the thigh or whatever. But was it?
This led me to think that the movement of the actual thigh-bone was actually not as extreme as the movement of the thigh itself. A small turn on the part of the thigh bone might influence the vertices which make up the thigh but will such a small turn move an object attached by it's hot-point?
This line of thought was what made me try out the external proxy objects in the wooden man I showed. (Before I threw in the towel!)
Marcus, you really had this concept figured out long ago, as evidenced by that vid. Really cool stuff!
I think the leap that Diomede came up with of having several different separated parts but all part of the same vertex object so it was as simple as unchecking self collisions is what really makes this puzzle solvable and doable. I've never been into animation before, still renders were always my thing, but now I can't stop playing with animations because this stuff is just so damn cool :)
Good point about the hotpoint of the bone vs the hotpoint of the soft underamor bodypart. Since it's all one mesh though, composed of several different separated parts, I would think it can't the hotpoint of any specific soft underarmor bodypart that plays into where it sits on the bone, since the method seems to work ok with different bones moving in different direction. I would guess it's the vertices being soft attached to the bone that take pre-eminence. Although as to the actual size/shape of the bone itself, that's a good point. We've just been taking it on faith that the bone run's the length of the area we need, but it's totally possible that's wrong, and might explain some measure of hijinks or variance.
I think the LaRoo example is very helpful, as is your brainstorming about rigging, and Joe's flying panels, and Roy's reports from Blender land and Carrara 8.1. I have been pondering whether some of the tedious under suit armor construction and soft body attach steps can be done in a plugin, which would depend on in part on the kind of questions you are pondering in your brainstorm. It seems to me that we have several mesh/rig setups to consider.
- old Poser cr2 like M4 and V4 have polygons of the mesh assigned to different named groups and then the rigging assigns those groups to bones (I think?).
- Carrara's native rigging does not require that the mesh first be made up of polygon groups, the skeleton is merely attached and the influence of the bone on any given vertex can be adjusted with weightmap tools.
- then there is the Genesis triax system with a protected mesh topology
- then there is the more recent Poser style weightmapping
The tests with V4 and M4 etc are based on that old Poser cr2 rigging system. When we do softbody attach for the panels, are they being attached to the bone? To the polygon group associated with the bone in the cr2 file? Proximity?
But Carrara's native rigging is based on a vertex, not on a polygon group. Creates questions in my mind that I can't answer, but that might affect the best way to construct and attach the soft body under suits, how vertices groups are assigned, etc.
To my mind, it is worth continuing to try different (a) base figure mesh groupings, if any, (b) different base figure rigging types, etc., etc., etc.
If it turns out that a cr2 based figure already has the core information required to create the undershirt, and to assign the the vertices groups in soft body attaches, then a big part of the tedious work might be able to be done with a utility, or maybe even a plugin. Wouldn't that be great! There are some amazing plugin creators. I would be willing to pay something for an automatic under suit creator and soft body attach utility, if the result allows for efficient custom dynamic cloth.
Heck, maybe Daz might even recognize the opportunity and include a function like that when they release Genesis3 figures for 3D printing and want Carrara users to pay for Genesis3 compatibility in Carrara.
Talk about wild speculation, but I think that brainstorming and multiple testing can help lead to innovation sometimes.
EDIT: Jon posted while I was typing. Stringtheory and Marcus deserve all the conceptual credit. I will accept some credit for testing and cheerleading, though.
Thanks Jonstark for your reply to the mixed-up ramblings of my mind.
Purely as a layman with no training in 3d my understanding is that the bone method of animation is that each bone has a number of vertices which it influences (weight-painting can change the vertices influenced and, I think, the degree of influence).
But I am supposing that attaching an object to a bone isn't the same thing because attaching isn't the same as rigging.
A parent-child attach to a bone is one thing and a soft-body-attach may or may not be another. GAHHH - my brain hurts.
But anyway, I looked at some failed simulations frame-by-frame and it looked for all the world that the cloth's weight was dislodging the sphere from the thigh and the turning effect of the bone was insufficient to give the sphere the force it needed to fight back.
That was why I tried out the external proxy objects. They were children of the bone and the collision spheres were soft-body-attached to them.
Another factor is that I understand that Carrara's rigging system is different to other systems such as Poser, Daz (and Blender??)
I think I'm bringing the conversation to the dark side where speculation and guessing is all that's left. (for me anyway).
I tried the parent child attempt first, earlier in the thread, and it works pretty well for a single bone and a single soft body part, but multiple body parts will crash into each other and shred or produce flying cloths (also for some weird reason I got the bug that I can do the animation ok in the assembly room but can't render it except via external renderer like Octane). Moving to soft body attach for each part seems to eliminate the problems I experienced with the parent/child method. I think the bone just grabs the vertices selected for the attach and holds them in their position relative to the bone (could even have a soft body attach to the bone that was no where near the bone and outside of the character, and it would still move in tandem with the bone I think, though I haven't really tested that yet).
Hi Diomede, your post crossed mine!
Regardless of all the mysterious inner-workings of physics simulations at least we're all enthused with the effectiveness of soft cloth in Carrara! I am totally sold on it for still renders (if only I made some, sometime). Although the Daz PAs make incredibly beautiful items, anyone using Carrara could make their own flags, medieval tents, washing lines, curtains, tablecloths, etc. with minimal modelling.
That is what impresses me about soft-body clothing. You only need a cylinder with its lower end flared out and it will drape nicely into folds and pleats.
You can soft-body-attach vertices here and there and they will stay where they are during the simulation and will cause variations in the drape. For example, if a wide neck opening has 24 vertices and every third one is attached, there will be a crenelated effect after the simulation. You can make a 'belt' with such vertices. Or stiff shoulders.
If the garment starts off as a disc-like shape, the figure can be posed doing (fairly) high-kicks - even though it looks kind of dubious to begin with.
Here is a video of some old renders I posted today. A couple are in there that I showed before:
https://www.youtube.com/watch?v=pcQewwMXv78&feature=youtu.be
Here are a couple of drapes in gif format.
I agree Marcus, it's heady stuff, I find myself watching animation of softbody cloth over and over, it's mesmerizing, and those are some great examples :)
Yes, bravo, Marcus. What you have done with the actual clothing articles is fantastic. I think we are very, very close to having some useful best practices for relatively quick construction and assigning of the undersuit, even for morphed figures. I can't wait to see what happens with the outer clothes that you have blazed the trail for. I love the scene you posted of the corner of the skirt set to the extended hand. I can picture animations of 18th and 19th century waltzes with women in long flowing gowns. I can't do them ;-) , but I can picture them and I'm sure folks will do them soon. :coolsmile: