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
Yes, as I mentioned long ago in my Bullet tutorial, the biggest challenge for a physics simulator is determining, from one time step to the next, which surfaces, if any, have moved from "outside" another surface to "inside" that surface.
The simulator breaks your "scene" time into "time steps" that define at what points in time the simulator calculates all the positions of the colliding objects. So if your scene time is 1 second, it will calculate maybe 100 times in that second, depending on the Accuracy setting.
And if the relative speeds of the two surfaces (let's say a cloth and a sphere) are too fast, then the change in distance between the two surfaces at subsequent calculation time steps might be extreme, and the simulator might get confused.
Therefore, the only way to cure that is to calculate more time steps, so you don't miss the point of collision. And that's what Accuracy does. And that's why it slows the sim down. Double the accuracy and you double the time steps, and the sim takes twice as long to complete.
If you can visualize a single vertex which is part of a softbody cloth mesh, approaching a single polygon of a rigid body sphere mesh. First the simulator has to decide what "inside" the sphere means, and what "outside" means. Same with the softbody object. With a single vertex approaching a single sphere polygon, you can start to see how challenging that is. And if at one time step the cloth vertex is way outside the sphere, and the next it's way inside the sphere, the sim can get very confused.
So either you slow the relative motion down, or increase the number of points in time when it makes the calculations.
Okay here's a slightly tweaked version of my sim using 7 pieces of underarmor. Forgive the knee joints, I just did a quick tweak to smooth the armor a bit. But really I think if you're going to use recycled character mesh for underarmor you will need to do some significant tweaking at the knees since the thigh and shin mesh don't meet at the knees. Annoying.
Also note the folds, where the cloth crosses itself since we can't use self collision in this sim. Annoying.
And I did add some more leg motion for Roygee. :) :) I was considering re-animating the character to do the Frug, but I lost interest and moved on to other things :) :) :)
And keep in mind, the reason the cloth is passing thru the hands and feet is because I removed those pieces of under armor. I suppose I could re-try it with a complete skeleton of underarmor, but I think the point is made
http://youtu.be/UKhRWUupFhU
BTW, it seems to me the biggest opportunity for refinement of the settings I've posted are in the Self Collision and Margin settings for the cloth object.
Unfortunately, those settings depend on other aspects of your scene, so there's no correct answer.
In this "SB attach w/ underarmor" sim you need a high Margin setting because the underarmor doesn't perfectly match the base character. So you need a high collsion distance to minimize poke thru.
But like I said, if you have a very dense mesh for your cloth (which you typically would), then a high margin, AND self collision enabled, will cause your cloth to explode, or at least get very rough as the sim tries to push adjacent vertices apart to match the collision distance.
You can solve that by disabling Self Collision when you have high Margin settings, but that has its own problems....like cloth falling thru itself.
So there may be, for your particular set up, a "sweet spot" of self collision and margin settings for the cloth.
For the sim I just posted, I was able to bring the margin down around 20% with only minor poke-thru. And if you can make those underneath body parts invisible to eliminate the poke thru then you may be okay.
I do enjoy a good leg motion - but it has to last for more than 0.03 seconds to get the full enjoyment out of it :)
More leg action for Roy...
http://youtu.be/iglheHTqspQ
That's an excellent looking test, Joe!
I want to also point out that the Joe's 3 posts at the top of this page and the 3 at the bottom of the prior page are filled with a ton of *excellent* info, so much so this is a part of the thread that is worthy of bookmarking all on it's own (and I have done so!). Some of this has been mentioned in other parts of the thread and other places as well by others and Joe himself, but I appreciate having a condensed 'spot' for a lot of great tips and strategies and theory altogether. Hence I'm bookmarking, and also wanted to say thanks.
Yes, Joe, thanks for posting your tests and observations with such detail. Very helpful. I hope someone gets a chance to repeat your last couple of videos with enclosed clothing meshes rather than loose sheets. Be curious if some of the issues with enclosed meshes that Roy mentioned are resolved.
This post is about some problems with underarmor construction. Because of blending, joint control morphs, and other features, the actual bending of the underlying mesh does not occur where the bodyparts meet. Therefore, there are some predictable problems with designing the undersuit to exactly match the base figure mesh, such as by exporting and reimporting the mesh.
Here is the problem illustrated at the shoulder for the Genesis 2 female with custom morphs. The undersuit panels match the chest, neck, collars, and shoulders. Each panel has been soft body attached to its corresponding bodypart. When the shoulder is rotated downward but the collar untouched, the base figure's collar vertices also bend downward to some degree in the blend zone but the bone fore the collar does not move. Because the collar panel is attached to the collar bone, the collar panel does not move downward when the shoulder is rotated. Therefore, the sleeve of the shirt does not bend downward. The result is that the sleeve of the shirt also fails to drape downward and the base figure's arm pokes through the bottom of the sleeve. This is certainly not new information, but I think it is a clear demonstration of what one basic problem around the joints.
As a result, I think the export/import approach might be a good shortcut for deriving the core of some panels but that it will have to be supplemented with deletes at the edges, and perhaps with spheres, capsules, or other ideas near where the mesh actually bends. I think something like a knee or an elbow that only fully bends in one direction might be OK, but shoulders (and other joints that have a wider range of movement) will require more thinking and more work.
I'm trying out a new method of handling the joint at the knees, where I closed the thigh bottom and extruded a few times to make a rounded bulb from the thigh, dragged it out to the side on the x axis in the modeling room then did the same for the top of the shin, so there's essentially two ball-like ends for both the thigh and the shin that are sitting in the same space (covering the knee but attached to different rotations).
I thought since these are sitting in the same space and intersecting each other that I would get lots of shredding or explosions or whatever, even though self-collision for the undersuit is unchecked. I remembered getting that kind of reaction before, but maybe I'm mis-remembering because my first test, which is to have my V4 walking forward for several cycles, I don't see any shredding, or if there is shredding or distortion it's very slight, and wouldn't push the cloth in any weird ways. So maybe this is a workable solution for joints? I'm thinking something similar could be done where the shoulder/collar meet, and if necessary the elbow region too, but that's just a guess.
Jon - good job, I think that is going to be the best way to deal with knees and probably other difficult joints.
Hi people. Just want to say, great thread. Dynamic strand hair and dynamic/ physics clothing on Genesis drives my decision to get on board Carrara.
I have succeeded in making some Nvidia PhysX soft hair and clothing via important static props for IClone 6 G6 characters so far. So no stranger to game-engine-type dynamic clothing. PhysX soft clothes (Apex Cloth) and Bullet soft physics have some overlapping areas. But Bullet in iClone/ Blender/ Modo isn't the same Bullet, so lots more to learn.
Based on my random mark-arounds in Carrara so far, I believe it is possible to have iClone 5 kind of "good enough" Bullet physics soft clothing in Carrara. Good enough, as in, consistent, simple enough to implement, fast enough to render. And, "won't look like rubber dress". Apex cloth is easier to learn and user-friendly, at least the way iClone dev designed it. Bullet cloth in Carrara is...for the long-attention span sorts.
But yes I'm still new to Carrara. Still ignoring the modelling room to focus on Genesis DS ZBrush pipeline. And have little concept of how pre-Genesis characters work differently. Info holes aplenty basically.
Like, even if I get soft clothes to work on Genesis, will the method also work on non-Genesis rigs? If I post my finding here, will I add more noise or signal?
The more I research on various soft physics implementations, the more I realize we're working with crazy amount of variables, proprietary weight issues, physics settings global and local, topological correctness...among others. So I stop my random experiments, now keeping it as systematic or "scientific" as possible. Narrowing down how Genesis Triax behaves in Carrara...
First:
- does it matter what kind of weight system the Actor Figure has when attempting to soft cloth onto them?
- Is V4 General or Triax or something else? If not, does V4 bone weight behave differently in Carrara compared to when in DS?
Basically, I'm wondering if most trials and errors reported here in this thread are V4-focus, so are they even applicable to Carrara-Genesis clothing?
Thanks, and keep the tests going. We're getting there...
Mythmaker, in the balance of this thread what you're seeing isn't cloth colliding against the character at all, but rather softbody cloth colliding against an (invisible) soft body undersuit that's 'soft - attached' to the bones of the figure.
The figure itself, whether V4, Genesis, Genesis2, or other, has collisions unchecked in the effects tab, so the same principle should work on any character, because for the purposes of the cloth calculations the character is a ghost :)
The reason for this is that while you can drape soft body cloth on a static unmoving figure, what was discovered early on is that if your character is animated and in motion, then the softbody will fall right through the character's mesh.
Stringtheory figured out the way around this is to have the softbody cloth collide with another softbody that's animated, and they will actually collide the way they are supposed to, even while animated.
This means that in effect you could use dynamic cloth on any character regardless of how it's weight painted, all that matters is the movement of the bones, so should work just as well on Genesis2 as V4.
The advantages are that the calculations for the cloth sim seem to be very fast and give good results. The downside is we're still trying to find the right settings for the various types of cloth effects, also there's a downside that in this method it takes some time to set up an initial invisible undersuit for the cloth to collide against (you can see my latest iteration in the screenshots in the post above this one, for example).
Actually while this sounds complicated to describe, all of this is fairly simple to do. I think Stringtheory said he would be putting out a tutorial on the method, but was delayed and not sure when it would be put together. I think a quick video tutorial would quickly demonstrate how this method works and how to put it all together. I'd volunteer to throw one together, but none of these concepts were my idea, I'm mostly just following paint-by-numbers in the footsteps of others in this thread who have pioneered the approach. Regardless, I'm very happy so far with the results and it's a lot of fun playing with cloth animations (especially when the cloth sims for me have been so fast as to nearly be realtime).
Here's a quick animation I rendered with the Octane for Carrara plugin, testing the concept of putting Carrara hair and fog primitive into an Octane render, to give one quick example
Hi Jonstark. I like visuals! And you do make some fine ones too. Will get to your wonderful dynamic hair videos once I pin down soft physics. Oh and good to know OR4C plays nice with Carrara dynamic hair. One more item on my to-buy list. Thanks for that...
Briefly...I marked around in Carrara "blind" before glossing over this mega-sized thread. I wasn't sure if V4 test applies for my Genesis case so avoided info-overload. Good thing is now I can confirm some of your joined findings without front-loaded prejudice. Also stumbled into my own blind alleys and alternative solutions.
Agree, soft body is not soft cloth, fabric setting is key. In fact, I see it as the first problem to solve. Pointless to test cloth draping without confirming it is fabric not thin latex sheets that I'm draping? :)
But there are more turning soft body into soft cloth challenges and solutions unique to Carrara-DS universe... Some of the "blind alleys" in this thread may be viable after all.
Actually, my limited understanding is bone weight should matter - especially when actor figure is set to non-collision. Not just weight type, but bone group and bone hierachy too, among other rigging related factors. The fav method in this thread, Soft attach undersuit bits to skeleton, affects bone weight, either canceling it or multiply it.
But perhaps bone weight is less consequential when only one or a pair of symmetrical collision mesh is used, as in your case of a skirt... I'm hoping for a solution that can work on ancient Chinese gown with flapping sleeves or modern smock dresses.
I swear I read somewhere (PhilW? Dev Spooky?) confirmed that TriaX matters in soft collision. Base on that I made a one piece super low-poly collision suit for Genesis. It kinda works, but still unstable, tearing etc.
I also question the topology of undersuits, one piece or chopped up. Especially the viability of panel style. Based on convention, Bullet or PhysX don't like anything with holes, or "concave-ness". The ideal collision shape is CONVEX.
I'm making a compound suit made up of convex capsule meshes. What I will call a CompoundSuit for clarity. Still experimenting.
Noted is your cylinder attach method: probably the most efficient collision object choice here in this thread, based on convention. Probably why you got good fast result with it. The downside is of course no tight fitting soft collision, good enough for frumpy clothes but not silky bodycon dresses...
Love to see where you take this with your collision mesh design...
Right now just to consolidate data to help myself from being brain-scrambled: (lol)
Key Variables = the relationship of Carrara soft body setting <---> bone weight <----> collision mesh design choice <---> attaching method. </p>
I'm getting a grip on the first. Now onto its relationship with the others...
Even at this point I can at least be certain that...even with soft physics settings "properly calibrated", other possibly more efficient "draping" method remain. I will post if and when I develop consistent result with alternative methods...
Thanks for the good brainstorming... Loving Carrara more the more I use it...
Mythmaker - good to have you on board. Just a little correction, Carrara Dynamic Hair does not render at all in Octane, Jon composited it in from a separate Carrara render pass.
Howdy PhilW. I subscribed to Infinite Skill on your tuts alone. Really helped.
Aww just when I thought Octane is finally rendering dynamic hair. Oh well. Thanks for the heads up.
EDIT:, wait, Phil, can you confirm TriaX figure weight does matter in collision mesh design in Carrara? I can't test that because I have no reference point for other rigs in Daz3D universe.
Howdy PhilW. I subscribed to Infinite Skill on your tuts alone. Really helped.
Aww just when I thought Octane is finally rendering dynamic hair. Oh well. Thanks for the heads up.
EDIT:, wait, Phil, can you confirm TriaX figure weight does matter in collision mesh design in Carrara? I can't test that because I have no reference point for other rigs in Daz3D universe.
Using the techniques in this thread, I can't see that it would make any difference. In this technique, there are no collisions with the mesh itself, hence the exact joint definitions wouldn't matter. Pieces of "under-armour" are soft-body-attached to the bones of the figure, and the cloth collisions are calculated against that (and only against that). So it should work with any figure with any joint type. I've not had the chance to do much extensive testing myself as I am very busy with a commercial animation project (using Carrara) but I have been following the discussion with much interest, as it resolves (or at least appears to resolve) one of the main areas that was missing in Carrara.
Has anyone had the softbody selection "named verteces" functions stop working? I went through and named all the verteces groups of an undersuit. The list of saved vertex groups appears and they are all listed, so I am sure I didn't name a polygon group instead of a vertex group. I can't select by vertex group. I have to use the brush. Argh.
It is a mystery.
I experienced something similar. I thought it was due to shading domains on the model because it worked after I deleted them. It could have been due to something else but I never investigated further.
Thanks for the quick reply. I don't have any shading domains. Seems pretty random.
I don't think I've had the exact thing you're experiencing. I've had my undersuit not 'remember' correctly which vertices are supposed to be a part of which group and get random verts attached to the wrong bones (which means the cloth sim goes out the window when I try to run it). My way around this is to save the undersuit as an obj in the objects tab, it seems to keep things straight when I do that; when I just save a scene with the undersuit it gets the verts mixed up every time, and I have to go back through and straighten them back out.
That can't be fun after all the setup work. I haven't had this particular issue. It could be the naming, some apps/sub-apps don't like spaces?
The only related glitch I experienced is sometimes clicking Edit or Select tree I get a long freeze, then I need to hit Esc to get back to the main Modifier tab then start allover again.
Using the techniques in this thread, I can't see that it would make any difference. In this technique, there are no collisions with the mesh itself, hence the exact joint definitions wouldn't matter. Pieces of "under-armour" are soft-body-attached to the bones of the figure, and the cloth collisions are calculated against that (and only against that). So it should work with any figure with any joint type. I've not had the chance to do much extensive testing myself as I am very busy with a commercial animation project (using Carrara) but I have been following the discussion with much interest, as it resolves (or at least appears to resolve) one of the main areas that was missing in Carrara.
Sorry Phil totally forgotten I asked a question here...
It was important for me to confirm if Carrara Bullet soft body play nice with TriAx conformed clothing (it doesn't). It will acknowledge the clothing figure's global Follower attribute but messes up skeletal relationship and weighting. This applies for soft conformed clothing on soft collision undersuit.
I recently discovered Sparrowhawke's Cloth Deformer. It came out a year ago out of a need to have higher degree of control for DS autofit/ TriAx-weighted/ Genesis conforming clothing. It takes into account the clothing figure's default bone weight, which can be fine-tuned further using 4 other types of weight map controls.
I have since concluded overall Cloth Deformer workflow is more efficient, and I can even get by without collision suit since Genesis is relatively low poly.
As for Bullet soft body dynamic clothing's usefulness for Genesis/TriAx clothing, it can be used for simple skirts, and tube top, preferable minimal limbs contact. And subjected to a highly specific collision figure and re-rigging method.
"Re-rigging" for Genesis or rigging for gen4 - because soft attached one by one to each bone, is essentially rigging from scratch. But without the benefit of proper weight map brushes. Rigging a fullbody or partial body multi-soft-attach undersuit (the current workable method) would be super challenging when the only weight assigning tool is per vertex with no fall-off.
The lack of fall-off control has apparently prompted Sparrowhawke's innovation which emphasizes paint brush friendly gradient fall-off.
Currently for Genesis clothing Bullet soft is just too much work for its worth in its current primitive form. Carrara Bullet soft can be made 10X easier with Poser's wonderful multiple weight type, each with highly nuanced weight brushes...
I have no experience with Poser dynamic cloth except video tutorials, enough to make me envy Poser users. But I think Sparrowhawke is close to bringing a simplified version of Poser dynamic cloth power into Carrara - via Daz Studio...
Cloth Deformer is not a true physics sim, its fabric does not super flappy like Bullet soft, but it does the most basic thing: enforced conformity with near-zero explosion rate.
I'm just glad to have clothing that collide with selective items in the scene...
But perhaps I shall not derail this gen4 specific topic...
My problems with the softbody attach undersuit have nothing to do with the physics calculations and everything to do with the saving/exporting functions and the ability to select by named vertices.
https://www.youtube.com/watch?v=YhIzj2N-St4&feature=youtu.be
This example has some very minor poke through that I will be able to fix easily, but the mesh of the undersuit lost the named vertices when I saved it to the browser, and the mesh of the undersuit actually changed when I simply exported it as an obj and reimported it. But anyway, I would put this in the success category.
I think a good short term goal would be to get the save function fixed for named vertices. That would go a long way toward being able to have a saved undersuit for a character that could be used for an entire wardrobe.
I experienced something similar. I thought it was due to shading domains on the model because it worked after I deleted them. It could have been due to something else but I never investigated further.
I just had this problem happened to me, first time. Thank goodness the scene in question is dead simple. In my case the problem is solved, at least in this one case.
Condition...
Empty new scene. Genesis loaded. GoZ to Zbrush. Made a hair object against Genesis. GoZ hair object (basically an OBJ) to Carrara. No group, no material, no shading domain. Just dumb 600 points single mesh.
In Carrara, I moved hair object into position. No parenting. Next, name vertices. only the mesh bits I wanted attach were named. The free verts left un-named.
Back in Assembly mode, add soft modifier (default + self-collision on 20%) and attach modifier to hair, attach to Genesis: Head. That Named Vertex Group was there as it should be, but when clicked, mystery...nothing happened. I was like, crappp...now it's happening to me too... Anyway, manually picked some vertices and proceeded with sim experiments.
[EDITED 2 hours after posting this]
Then I decided the existing attached vertices weren't enough. Also decided to make new attempt at named vertex group. Went into Modeling mode, added a few more vertices to the existing Named Vertex Group, went back to Assembly mode.
Guess what... the Name Vertices mysteriously work now. Red dots exactly where I want them to be.
So it seems, the named vertices need to be "re-activated" by editing the existing vertex group, by adding new vertices, or probably removing some. Kind of like, wake it up.
But our test scenes are all different. I hope this works out for your case, Diomede64.
Good luck...
Found the named vertex group bug solution I think... See edited last post.
It's been almost two months since this discussion was active - is anyone still playing with this or using it? Are you finding stable solutions, or is it all proving too difficult?
I have been playing with it again today for a bit, and have produced a stable cloth simulation - just raising and lowering the leg with a skirt, but it is better than I had previously achieved! I am still getting a bug where the Carrara renderer is not advancing the cloth frames, but by rendering in Octane I have got around this (amazing that this is something that works in Octane and NOT in native Carrara!).
I'll try to get it posted shortly, but just wondering if anyone else is still playing?!
Hi Phil, yes, I am using it for my full figure shots that need believable cloth dynamics (it is absolutly usable). However I am only using it for skirts, capes and flowing sleeves. The problem with tight clothing is that the undercloth is in constant "wiggle" motion which tends to cause tight outer clothing to continuously move and wrinkle even when the figure is not in motion (setting the undercloth margin to zero helps quite a lot though). For my background figures I have had a lot of success creating soft cloth morph sets using my undercloth rig. The results are surprisingly realistic and resource efficient. My end product will contain a combination of real dynamic cloth, morphed cloth and conformed clothing.
I haven't abandoned it but I have not been using it lately. More to do with the nature of my recent projects than anything related to the method.
My biggest complaint is that after I name the vertex groups within an undersuit, I cannot seem to preserve the vertex groups correctly when saving and retrieving the object to the browser. Therefore, I have to regroup every time I load an undersuit. Draping in Poser and then importing the obj can often be easier for me.
However, on a case-by-case basis, I definitely still use the concept. I just have nothing new to report.
diomede64, do you still have the named verticies issue if you save the entire scene and then reopen as opposed to saving the figure to the browser?
Thanks for the feedback. I tried a walk cycle with my setup, it got to around three seconds and then the skirt exploded for no apparent reason, so it is still not as stable as I would like.