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
Oh, yeah...Geometric Fidelity...I think that's the one that is only used for rigid body solutions, not soft or cloth solutions. But it's been a while since I played with it.
You may be right about geometric fidelity. I'll run some simulations and see if changing the value makes any difference.
I believe the intent of geometric fidelity is that the simulation replaces a rigid physics object with a simplified model in order to speed calculations and collision detection. The simpler the object the better. So if you can replace a human shaped object with a simple cube you speed up the calculations immensely. But if you want the object to be closer in shape to the original object, you have it use more complex objects, hence more geometric fidelity to the original object.
I originally thought it might apply to the object that cloth is colliding with also, but after a bunch of testing I think I found out that it makes no difference.
I think. But my memory isn't all that hot lately.
That's really good to know. There are so many of these setting that could be doing any number of things. The labels don't always make it clear what they are effecing. It would have been nice if all of them had tool tips like some do.
So I made V4 the source object of a surface replicator and then covered her with little soft body polygons. Not a feasible approach. lol
Hi folks,
It's great to see that this is progressing nicely...
Diomede & Evilproducer, thanks for your generous remarks but Stringtheory is leading the way here - his explanation of code-bases under the hood is fascinating and it leads me to think that DAZ gave us more than perhaps we knew. (If I took a bow I would likely manage to hit my head on something...)
Having only low-spec hardware myself, I can understand the frustration for you, Evilproducer,of not having the latest version of Carrara but you're taking a lively part in the discussion - as you did in my earlier thread - and it helps everyone very much. I hope you get to play with it sometime soon.
If I take a leap ahead in thinking of what the final outcome of all these experiments could be, I would hope for collision-clad figures and some basic styles of clothing which would drape and move along in animations without poke-through. It would be a great enhancement to Carrara.
But I think that we will never get any one set of final settings for all the variables that are available.
I believe that some settings are affected by the size of the object in the scene. Other settings depend on the density of the mesh and the way it has been put together. What I'm saying is that I believe that individuals will need to tweak their own settings for each project. But it is really nice to see a drape turn out well and it will be even more satisfying to see animations work well.
Stringtheory, I haven't re-visited any of the experiments I tried at the start of the year or even read through the things I posted but I think I can identify where I went wrong.
First I was creating a piece of clothing to drape as a soft body item.
Then I was also creating stretched-out spheres to act as collision objects. It was my understanding that by adding a soft-body-attach modifier to each sphere and, by including every vertex in the sphere, I was creating (pseudo) hard-body objects.
These 'hard' spheres collided with the cloth to some extent but always failed in animations. They were good for one hit only.
I think that what you are doing differently is to FIRST make your collision cylinder or sleeve into a soft body by applying the soft-body modifier. THEN you are selecting all the vertices as soft-body-attach. (You have them selected as a named vertex group).
I think missing out the first step of making my collision spheres soft cloth was the reason I failed. But that is from memory - I might be wrong.
Carrara allows a soft-body-attach modifier to be applied to a mesh without first applying a basic soft-body modifier and, without going on and on, I think the function has its uses.
Ok get ready for some really stupid questions :) I've never ever done a bullet physics simulation for anything before, but just finished some rendering work and want to play with this approach. Stringtheory, I really appreciate you posted the settings for the 'under body' soft body object parented to the V4/M4:
Can anyone give an idea though of what the settings should be for the actual cloth softbody that will be draping over the underbody object? Not sure exactly how to make clothlike softbody.
Also is it necessary for the underbody and cloth objects to have thickness, or is it ok for them to be planar? Maybe it's even required to be planar...
When you put the 'underbody' object on the character, are you simply dragging and dropping it on the relevant body part in the scene tree? I guess I'm asking the method of making the 'underbody' parent itself to the character in such a way that the character movements that are keyframed make the underbody object move with the character without there being any keyframing involved.
Sorry these are probably super obvious questions, but I've never done any playing around with this stuff before.
Oh yeah, additional dumb questions, does the 'cloth' soft object attach to the 'underclothes' soft object? Is that how it stays on without slipping off? Should be attached everywhere, or just at a few spots?
I am getting closer, but still need to learn a lot more. However, after a few experiments, I join Stringtheory in believing this is a concept that can be used for some purposes.
- I loaded Aiko 3 and modeled a vertex object with subsections roughly corresponding to the bones (I wish I had been more exact).
- I named the vertices of each subsection
- I applied a general soft body modifier to the vertex object with the settings Stringtheory suggested
- I applied a softbody attach modifier for each subsection/bone combination, using the select named vertices option under edit
- for Aiko 3, after selecting "model", I unchecked collide with other objects
- under scene, I used the settings Stringtheory suggested
- I did a test by advancing the keyframe at to 2 seconds and posing Aiko3
- Clicking the simulate physics icon resulted in the undercloth subpieces moved with the bones (but because I was sloppy, there remained imperfections)
- I returned to the starting pose / origin keyframe
- I inserted a new vertex object and modeled a simple long skirt
- I applied a soft body modifier to the new skirt
- I clicked simulate physics. The simulation took 3 minutes. The skirt moved as expected when the legs moved.
- I unchecked visibility for the undercloth object
- Note - because I was sloppy with the undercloth pieces, there remained some poke through.
However, I think this is confirmation that the general approach of disabling collision for the figure (Aiko3 in this case) and using soft body attach pieces as a proxy can save significant calculation time and generate workable results. Now to try experiment with various settings.
@John - I'm not sure what settings should be used for the outer cloth. I just used something similar to what Stringtheory had suggested for the undercloth. I did not add thickness, but it would be worth testing.
Note - I can also envision this use of a proxy and softbody attach for something like a necklace, with a panel placed on the neck and chest.
Great work Stringtheory.
this is what the modifier tab looks like for the vertex object being used as the proxy. I mean the undercloth. Note 8 different soft body attach, one for each set of named vertices, in addition to the general soft body modifier.
Here is the same skirt soft body draped on Aiko3 in the same pose without a proxy (no undercloth) and just colliding on Aiko 3. It took 12 minutes to calculate, so 4 times as long, and still has problems where the back leg stretched the skirt. Calculation efficiency seems to be a big difference.
A lot of the efficiency from using any proxy is that it contains far fewer polygons that a full figure, and so needs to calculate far fewer potential interactions. From my own tests I would say that Stringtheory is definitely onto something, and once a figure is set up with the underclothes, you should be able to use any dynamic clothing, including (hopefully) those made for Poser's dynamics - that would be great. There will probably be a fair bit of tweaking to get a perfect result, but at least the calculations are not too long now so the cycle of tweak-test-tweak will be shortened. I have found the issues tend to be at the joints such as the shoulders and knees, these may be difficult to get looking right, but not impossible.
Congrats on all the testing and the intellectual leap forward!
Great work diomede64!
The mesh that you modeled looks perfect. I have no modeling skills so I have wasted a lot of time trying to get a decent undercloth mesh.
The settings for the top mesh depend entirely on the geometry of the outer cloth. Here are some things to keep in mind for those settings.
Stiffness: This is the stretchyness of the polygons. If your mesh is fine and this is set low the cloth will stretch like spandex. Raise the value to make the cloth stretch less.
Bending: This controls the tightness of the springs that connect each polygon the to the next. Lower this value to achieve finer wrinkling.
But since we are all figuring this one out together the "best" settings are still being discovered.
I think the issue you are running into with your skirt is that it is just way too tight at the bottom. You could drop the stiffness value to give it a little more stretch or you could widen the mesh a little. In real life if that fabric was stiff denim the legs would not be able to open that wide.
I had a long busy day today but just wanted to drop in quick and comment on your work.
Jonstark, sorry I didn't get a chance to get to your questions sooner but follow diomede64 work. It is a great step by step example of the technique.
Speaking of techniques the screen capture below is what I was doing last night; surface replicating a single soft body polygon with distribution mapping onto V4. My hope was that this would do away with the need for a unique undercloth mesh for each different figure and body morph. Amazingly this kinda works but it takes a long time to calculate. unfortunately I think this one is a dead end so it's back to struggling to model a decent undercloth mesh.
Jon, if you are only starting out with this, I'd suggest taking five minutes to try a basic abstract simulation before complicating it with clothing on a figure.
1 In the Vertex Modeller create a default grid
2 Go to: model / tesselate / mid-edge to center
3 Select one of the faces near the middle of this cloth-to-be
4 Choose the extrude tool with normal extrusion to pull that face down 7 or eight times
What you should have is a one-legged version of the 'square pants' I used earlier!
In the Assembly Room, select this object and apply a soft-body modifier.
I'd suggest dropping the stiffness setting to 30% and the folding to zero. Also tick the self-collision box. If you tick the margin box, drop the value down from default for the first test.
Now choose soft-body attach. Click the edit button and paint the vertices of the 'leg' part of your object. You can lower the brush size and can choose + or - at the top left of the screen to paint or un-paint. These painted vertices will not drape - the draping parts of the object should collide with them. Stringtheory suggests selecting the vertices you would otherwise paint (while in the vertex room) and naming them. That is a great method to use because the soft-body-attach edit provides a button to choose named vertices.
Normally you would attach your painted or named vertices to something - such as a figure's bone or a flagpole or whatever. In this case just leave it.
Try a two-second simulation. The plane should drape down over the column and collide with it. The column should remain stiff. If an object sizzles or explodes during a simulation then the vertices are too dense.
Next simulation: Go for three seconds and rotate the object 45 degrees or so along the timeline.
Unfortunately I'm tied up for the rest of the day. I hope this helps although it hasn't addressed your exact questions. It may seem a lot to do from reading it but it should only take 5-10 minutes of your time.
Diomede, You're results for a first try are very promising. The suit you created already is impressive. Earlier in the thread, Stringtheory helped me with the issue of overlapping at the hem - he suggested settings which worked. Perhaps though,in your case, it is more to do with the gap at the knees.
As a general remark, I like to drop the stiffness setting much lower than the default to begin with and set folding to zero. I also tend to create 'clothing that has greater clearance from the figure's legs. That was to allow a wider range of static poses when that was all I was doing and isn't required as much now. But the tighter and more realistic a garment is to begin with, the less it may be able to form folds and creases during the simulation.
Another thing to try is to vary the mesh. If. from mid-thigh to hem, the polygons are larger (less subdivided) the effect can be interesting.
In the time I took to send my previous post, Stringtheory replied!
The stiffness and bending settings are IMO the major ones to play with. I was giving opposite advice to Stringtheory's but the settings all depend on the mesh.
Thank you guys for the tips and advice. I already started tinkering, learning quickly that I don't know much about what I'm doing :)
I first tried some soft body sims just inserting a cylinder vertex mesh that I made really skinny and a vertex plane that I tessellated several times. I was making a flag pole, and through trial and error learned how to select the vertexes on one side of the plane and attach them to the flag pole, then started doing simulations and experimenting with the various sliders. I haven't really got a super convincing cloth setting yet, found when I took the stiffness down low and the bending down super super low (like .2) that I could get some nice falling flag wrinkles and behavior, but it never seemed to 'settle' and seemed to want to continue to wave/bounce.
I then loaded V4, put a bodysuit on her, turned the bodysuit to a soft body, and took a vertex cyclinder, deleted the top and bottom to make it hollow, and maneuvered and scaled it so it would sit just around V4's forearm (imitating your first proof of concept in this thread, stringtheory). I added softbody to the cylinder, attached it to the the bodysuit, hit the physics sim button and... lol I'm betting you can guess what happened next. In extremely slow slow motion everything exploded (I obviously understood the concept wrong) And the sim locked up with slowness, even made the rest of my computer programs slow to a crawl :) I was going to wait it out, but thought life is too short, and I crashed out of Carrara after 20 minutes it was still less than 3% calculated.
That's what I get for skipping steps I guess :)
I next loaded V4 up again in default position, took another vertex cyclinder and put it over her forearm again, and this time took some time to actually model in the assembly room (even though I suck at modeling it's astonishing what a few minutes of soft-selected pushing/pulling/scaling of a few selected vertexes can do and I was able to fairly quickly make that cylinder into a form-fitting fore arm cover... what's it called in real life, bracer?). I was inordinately pleased with this, because I hadn't expected this to be so easy.
Anyhow, I gave this shaped-cylinder a name, dropped it on the V4 forearm in the scene tree hierarchy. Then I duplicated this cylinder and scaled the new one so it was wider, named this one 'cloth' and dragged it out of the V4 hierarchy so it was sitting in the scene by itself, even though it was positioned around the 'under clothes forearm bracer' I had made and just slightly bigger. I made it a softbody, and attached it just at the loop of vertexes right near the elbow to the 'underclothes bracer' I had made, then keyframed some arm swings and movements with V4.
I also turned off collisions for V4 (in the effects tab when V4's 'model' is selected in scene tree, for other modeling-inept people like me who might not know where to find this).
This time it calculated physics super fast, like only a few seconds, and I got some nice effects, the cloth didn't slip off (though since it was only attached at one end I found if I raised V4's arm above her head the end of the sleeve near the wrist would slide backward along the forearm, wrinkling a bit, not unlike what happens with a real sleeve I guess, so not a bad thing). There was no poke through, and it all looked pretty good, with the exception that because I haven't honed my cloth settings yet it looked a bit like a combination of jello and cloth rather than the cloth look I was hoping for. But importantly, no poke through, and a fast sim to show me that it was working :)
Now more questions...
Diomede, I'm certain you're doing it right, but I'm not sure that I understand the need for all the different soft attaches on the skirt. I would think there would be just one soft attach at the waist, and everything hangs from that, and won't the skirt naturally as part of the physics sim automatically collide/rebound off of all other soft body objects in the scene? Or am I drastically misunderstanding. I haven't progressed to the point of putting a bunch of 'underclothes' objects on the legs and seeing what happens with more than one soft body object able to hit the cloth.
Also I'm wondering if there's any ideas how we handle joints, particularly elbows and knees. I'm not sure how we can make softbody underclothes that will stay in the right place with something as dynamic as a bending knee or elbow...
Oh man this is fun. Nearly 4am and I can't stop playing....
Hmm, now my simulation animation in the assembly room looks great, but when I try to render my cloth object stays hanging static in the same position, not moving, not dynamic, nothing. Looks like I'm running into the same thing PhilW mentioned earlier in the thread. What am I doing wrong? Is there some render setting I'm missing or something?
Thanks for the feedback and the suggestions.
@PhlW
Joints are going to be a challenge. Just brainstorming, it seems to me that a ring of vertices could be named around the core of a joint and then a sphere could be softbody attached at that location. Not sure it would work, but will do more tests when I get a chance.
I know what you mean! I had thought of my strategy for creating the proxy mesh for Aiko3 as soon as Stringtheory posted the concept, but didn't have time during the week to try anything detailed. Then Friday night I couldn't go to sleep until I got at least one promising simulation with a long skirt. I have more ideas I want to try but it may be a few more days until I get a chance.
Thanks for pointing this out. In my reply to diomede64, the sentence at the end of the stiffness setting exlanation says to raise the stiffness setting to make the cloth stretch less. While this was intended more as an example of how the setting effects the mesh behavior, it did come across as though i was providing a solution to the the poke through he was experiencing on the skirt. So as to not cause any confusion for other following this thread, in diomede64s's case, lowering the stiffness would yield better results.
I wanted to check to see if it was the lower polygon count of the proxy mesh that was the cause of the time saving. I don't think it is. I think Stringtheory is correct that softbody collision with softbody is calculated faster than keyframe collision with siftbody even if the polycount is reduced.
I checked by creating a quick low-poly conforming bodysuit for Aiko3. I then turned off the collision for Aiko3 but kept the collision on for the bodysuit. Thus, the softbody dress would collide with the low-poly keyframed bodysuit figure that would closely track the pose of the main A3 figure.
The result took over 30 minutes to get just 40% of the calculations. At that point I interrupted it.
Therefore, I will continue to focus on finding ways for soft-body attach to create the collision proxy.
Hi PhilW, Did you find the cause of this. It sounds like a bug. I haven't come across this myself and wonder if it is because I'm on a Mac which uses the QuickTime codec. I'm asking because JonStark is running into the same problem.
I'm wondering what the outcome is when the render with this issue is saved to file (is the cloth still motionless) and also what happens if the render is rendered as individual files.
In my test, the undercloth is a single vertex object but cut in 8 sections. Each section is softbody attached separately to one of Aiko3's bones, not to the skirt (which did not exist when I attached the undercloth pieces). I'm not sure if I had to do each bone individually, but it was a convenient way to make sure that I attached each piece to Aiko3 in the correct location. The proxy mesh is not attached to the skirt. The skirt is separate vertex object. In my simulation, I did not soft-body attach the skirt to anything at all. Nor is anything soft body attached to the skirt. That skirt was a stand alone soft body. It was just a test. In practice, I am sure the top ring of a skirt would be soft body attached like a belt. There are examples earlier in the thread. One reason to keep the outer cloth mesh separate from the proxy is that it is simple to uncheck visibility for the proxy undercloth.
Ah, I see! Thank you Diomede, that helps the lightbulb go on a bit :) I sort of had the right idea about how the physics needs to work, just the wrong idea about which item was having the multiple soft attaches. This leads me to a new question though, which is do all the 'underclothes' need to be part of the same softbody object? I ask because originally I didn't realize that for example the bands on the leg in the skirt video that Stringtheory did were all part of the same object, and I kind of mentally was expecting to take a couple of separate cylinders and planes etc and shape them over various body parts (thighs, buttocks, etc), dropping each one on the relevant V4 body part/bone and soft attaching that object only to that body part. Is that approach of having several different underclothes parts destined for failure? I haven't got so far along as to try it and see, and was thinking maybe there's some flaw in doing it as separate vertex objects that I hadn't foreseen.
Oh and seriously guys, I can't figure out what I'm doing wrong on render settings or scene settings or whatever that's causing me to get beautiful physics sims in the main assembly room, but when I render the softbody object doesn't move at all, and so I have my V4 just waving her arm around, passing through a softbody structure. I know I'm missing something since you guys have attached some nice animations to this thread. What am I doing wrong? :)
Ah, I see! Thank you Diomede, that helps the lightbulb go on a bit :) I sort of had the right idea about how the physics needs to work, just the wrong idea about which item was having the multiple soft attaches. This leads me to a new question though, which is do all the 'underclothes' need to be part of the same softbody object? I ask because originally I didn't realize that for example the bands on the leg in the skirt video that Stringtheory did were all part of the same object, and I kind of mentally was expecting to take a couple of separate cylinders and planes etc and shape them over various body parts (thighs, buttocks, etc), dropping each one on the relevant V4 body part/bone and soft attaching that object only to that body part. Is that approach of having several different underclothes parts destined for failure? I haven't got so far along as to try it and see, and was thinking maybe there's some flaw in doing it as separate vertex objects that I hadn't foreseen.
Jonstark, no, the soft body attached underclothes do not need to be part of the same soft body object. This is also true for the soft body clothing.
However at this point I don't know what, if any, the performance differences would be between one soft body undercloth split into multiple pieces versus multiple seperate soft body objects, especially if the settings are slightly different on each one. This might make the math much more complex (just as having multiple lights do in rendering). We may also find there is no difference at all. My hunch is that there will be some penalty but it might be so minor that its a non issue. It's something worth testing and reporting on.
I don't think you are doing anything wrong. This might be a bug. Couple of questions: If you go to the middle of your animation and render a single frame, does the soft body match the frame in the assembly room or is it still the same as frame 0?
What happens if you render with a different renderer (like non photo realistic)? Also, what happens if you drag the soft body object a little in any direction and then try the render? Last thing to try, what happens if you export the soft body object (as a .car file) and then re import it in with the imported object in place of the original?
Hi PhilW, Did you find the cause of this. It sounds like a bug. I haven't come across this myself and wonder if it is because I'm on a Mac which uses the QuickTime codec. I'm asking because JonStark is running into the same problem.
I'm wondering what the outcome is when the render with this issue is saved to file (is tge cloth still motionless) and also what happens if the render is rendered as individual files.
Hmmm. I'm not having this problem. I have used both regular C8.5 Pro and the new Beta. Windows 8.1. I have used the default render settings. I didn't even try to render until this problem was reported. I had just posted assemble room screenshots. But I have tested it now. No problem. Hmmm. Can't even think of a candidate for the cause.
I don't think you are doing anything wrong. This might be a bug. Couple of questions: If you go to the middle of your animation and render a single frame, does the soft body match the frame in the assembly room or is it still the same as frame 0?
What happens if you render with a different renderer (like non photo realistic)? Also, what happens if you drag the soft body object a little in any direction and then try the render? Last thing to try, what happens if you export the soft body object (as a .car file) and then re import it in with the imported object in place of the original?
Tried rendering a single frame in the middle and it looked fine, the softbody was in the right place, draping over the arm nicely. So I thought maybe I was rendering as the wrong movie type or something, but no matter which format of animation I choose I get only V4's arm moving, and the softbody cloth is static. Interestingly when I tried rendering a movie format starting not at the beginning but deeper within the animation, it always renders the first frame exactly right, where the cloth 'should' theoretically be according to the sim, but the cloth doesn't move from that point and is totally static throughout the rest of the scene.
When you talk about exporting the soft body cloth, what steps would be involved in that? Do you drag it to the 'my objects tab, delete the original, then drag it back out of the objects tab into the scene and drop it back on the V4 bone/bodypart in question? Or is this a full export outside of Carrara and a re-opening of the scene? Also I'm just doing with this with the cloth, or am I doing it with the V4 and the underclothes vertex object too? Sorry this is probably obvious but I don't want to mess it up.
Also tried saving the cloth obj in the objects tab, then dropped it back into the scene. It did appear the dynamic sim was baked into it, as again it looked great in animation in the assembly room, but same problem persists in the actual render room. Also tried rendering with nonphotorealistic engine, problem remains. Guess I'm just lucky on this one :(
This is super puzzling. Especially the test with the baked sim. I can't duplicate the problem so it's hard to come up with a workaround.
What happens if you render as sequenced JPEG's?
Tried sequenced jpeg same as the other animation types, problem persists, unfortunately. Really doesn't make sense why it would render any particular static jpeg correctly, but doing it altogether at once the cloth sit's static.
I don't do a lot of animation's anyway, and this technique can definitely work to let me put dynamic clothing on my static renders, so this technique is definitely still very useful to me, but heck I really wanted to do a few 'v4 walks in dynamic skirt' type animations :) I'll keep trying stuff to see whether I can figure out what I'm doing wrong.
If I were to guess I would say that the sequencing mechanism for that object channel is getting corrupted some how.
I am at a lose on this one. I am curious though what would happen if you opened the baked object in a new project window and rendered.