Jim Burton opened this issue on Nov 10, 2005 · 57 posts
Jim Burton posted Thu, 10 November 2005 at 12:06 PM
zollster posted Thu, 10 November 2005 at 12:18 PM
i was rendering a head shot of 1 figure, 1 hair prop and three lights with just raytrace turned on. render got stuck 1/2 way through
Spanki posted Thu, 10 November 2005 at 1:29 PM
Jim, it looks like you have some reversed normals there, or some duplicate polys, or both.
Cinema4D Plugins (Home of Riptide, Riptide Pro, Undertow, Morph Mill, KyamaSlide and I/Ogre plugins) Poser products Freelance Modelling, Poser Rigging, UV-mapping work for hire.
DominiqueB posted Thu, 10 November 2005 at 2:06 PM
I ws wondering if they had fixed that in SR2...I guess not!
Dominique Digital Cats Media
stewer posted Thu, 10 November 2005 at 2:12 PM
Looks like reversed normals to me as well. Did you try checking "Normals_forward" on the material?
Jim Burton posted Thu, 10 November 2005 at 3:29 PM
Nope, it is a gen-u-wine bug. It shows up in some of the clothing included with P6, too. It doesn't show up with Poser 4, PP and 5. I don't even have normals in any of my mesh. ;-) Here is an eariler thread on it: http://www.renderosity.com/messages.ez?Form.ShowMessage=2301883
nerd posted Thu, 10 November 2005 at 8:02 PM Forum Moderator
byAnton posted Thu, 10 November 2005 at 11:37 PM
Jim, You might have am irregular polygon or vertex with no surface. That can prevent welding.
-Anton, creator of Apollo Maximus
"Conviction without truth is denial; Denial in the
face of truth is concealment."
EdW posted Fri, 11 November 2005 at 2:06 AM
Well I tend to agree with Jim on this one... this is a Poser bug. I've seen it on more than one outfit. I've checked the groups both in and out of Poser and the faces are pointed the right way. Poser is simply not welding them. If I take the same clothing pieces that have a problem in P6 into P5 or PPP the problem isn't there. The points are welded.
Spanki posted Fri, 11 November 2005 at 2:25 AM
I'd almost lay money on it not being a Poser bug... I betcha there's a backward poly, or an extra poly or a strangley formed extra poly in those spots. Send me the .obj file and I'll find the problem with it :) - seriously, I'm curious myself... if you can IM me download link to it (or a similar one that shows the same problem), I'd be happy to look into it and/or prove myself wrong.
Cinema4D Plugins (Home of Riptide, Riptide Pro, Undertow, Morph Mill, KyamaSlide and I/Ogre plugins) Poser products Freelance Modelling, Poser Rigging, UV-mapping work for hire.
Spanki posted Fri, 11 November 2005 at 2:28 AM
...just to clarify one thing... you don't need any 'normals' in your .obj file for the (Poser-computed) normals to be facing the wrong way. It depends on the winding order of the polygons. A polygon that connects to points: 1,2,3,4 will face the oposite direction as one that connects to those same points, but in: 4,3,2,1 order.
Cinema4D Plugins (Home of Riptide, Riptide Pro, Undertow, Morph Mill, KyamaSlide and I/Ogre plugins) Poser products Freelance Modelling, Poser Rigging, UV-mapping work for hire.
nerd posted Fri, 11 November 2005 at 2:47 AM Forum Moderator
Yeah, I'd like to examine that OBJ too. I've had things just like this and It always turned out to be a messed up facet or something else wonky in the mesh. It does usually seem to hit the edges though. I usually blame the freebie habware obj2max plugin. There's no way I could ever mess up a mesh ;-) When I saw these they would vanish if I switched to SreeD diaplay, the same diaplay engine P4 and P5 use. BUT, there is a bug I've encountered where body parts would come unwelded after a long session in Poser. It usually effected the neck or shoulders. Except that bug caused the entire seam to break, not just the corners. Nerd3D
EnglishBob posted Fri, 11 November 2005 at 4:56 AM
byAnton posted Fri, 11 November 2005 at 6:08 AM
Poser also crates a line of duplicate vertexes along the seems lines when it exports a figure mesh. Something wierd but when it exports it welds but preserves a border of vertexes which is why morphs explode and reodering occurs.
-Anton, creator of Apollo Maximus
"Conviction without truth is denial; Denial in the
face of truth is concealment."
Jim Burton posted Fri, 11 November 2005 at 8:38 AM
Jim Burton posted Fri, 11 November 2005 at 9:05 AM
Spanki posted Fri, 11 November 2005 at 10:15 AM
Uhm... but, that's a 'prop', not a figure, so Poser won't weld those seams. Those seams exist in the .obj file, as it is on the disk. If you 'export' it as an .obj, thse seams still won't get welded, because it's a prop, not a figure.
Cinema4D Plugins (Home of Riptide, Riptide Pro, Undertow, Morph Mill, KyamaSlide and I/Ogre plugins) Poser products Freelance Modelling, Poser Rigging, UV-mapping work for hire.
Spanki posted Fri, 11 November 2005 at 10:18 AM
EB, if you still have the mesh lying around, I'd be interested in looking at it... I'm still looking for examples of this problem.
Cinema4D Plugins (Home of Riptide, Riptide Pro, Undertow, Morph Mill, KyamaSlide and I/Ogre plugins) Poser products Freelance Modelling, Poser Rigging, UV-mapping work for hire.
Spanki posted Fri, 11 November 2005 at 10:24 AM
Back to the seam issue in general... my .obj files have no seams :). They don't need to. People always refer to 'cutting' the groups, but the groups don't need to actually be separate meshes. I guess one issue related to this is the software used to create the groupings... it might be cutting/splitting the mesh for you. I used C4D for my modelling, with a set of plugins I wrote to help me deal with and import/export welded, grouped meshes (it even retains and/or creates UVMapper 'regions').
Cinema4D Plugins (Home of Riptide, Riptide Pro, Undertow, Morph Mill, KyamaSlide and I/Ogre plugins) Poser products Freelance Modelling, Poser Rigging, UV-mapping work for hire.
EnglishBob posted Fri, 11 November 2005 at 10:29 AM
@Spanki: Did you just hit on something there, I wonder? Now that conforming clothing lives in the Props folder in Poser 6, I wonder if they somehow broke the welding action when those changes were made? I'll IM you for an address.
EnglishBob posted Fri, 11 November 2005 at 10:35 AM
My OBJ file has no seams either - it's all one continuous mesh. However Poser breaks the geometry into a separate mesh for each group internally, as you can show by exporting it again (without checking "weld", obviously). The split illustrated in my image is taking place at what was a single vertex in my mesh.
who3d posted Fri, 11 November 2005 at 10:43 AM
..
Spanki posted Fri, 11 November 2005 at 10:49 AM
Hmm.. you know, I assumed it was a prop, since it was in the props folder, but I see now that it is in fact a figure... interesting.
Cinema4D Plugins (Home of Riptide, Riptide Pro, Undertow, Morph Mill, KyamaSlide and I/Ogre plugins) Poser products Freelance Modelling, Poser Rigging, UV-mapping work for hire.
Spanki posted Fri, 11 November 2005 at 11:06 AM
After looking at this further, and possibly not related to the problem, the figure file for the bikini bottom (normally I'd call it a .cr2, but in this case it's a .pp2) is not really done correctly... - it has weld statements for non-existant mesh parts. - it has ghost joints for parts that can't possibly affect the buttocks or hip parts. - it has morph channels for those ghost parts which have no geometry to begin with. - those ghost parts (rToe, for example) reference other parts that don't even have joints set up for them (rBigToe1, for example). - it has 'storageOffset' and 'geomHandlerGeom' statements (top of the file) for non-exstant mesh parts. ...frankly, I'm surprised it works at all :). I guess Poser is more tolerant than I thought. Lastly, I also note that it is set up as a figure, but not a conforming figure, so you'd have to parent it to a figure and then apply poses to it along with the parent.
Cinema4D Plugins (Home of Riptide, Riptide Pro, Undertow, Morph Mill, KyamaSlide and I/Ogre plugins) Poser products Freelance Modelling, Poser Rigging, UV-mapping work for hire.
Letterworks posted Fri, 11 November 2005 at 11:40 AM
Spanki posted Fri, 11 November 2005 at 11:44 AM
Yeah Mike, that sounds plausable... I found no other problem with the mesh to account for the problem. It does sound like P6 treats that situation differently than P4 and P5.
Cinema4D Plugins (Home of Riptide, Riptide Pro, Undertow, Morph Mill, KyamaSlide and I/Ogre plugins) Poser products Freelance Modelling, Poser Rigging, UV-mapping work for hire.
Spanki posted Fri, 11 November 2005 at 11:48 AM
Jim, in your case (image in post #1), I'd move the split further up, where the thin band starts... in fact, I'd make the thigh parts a single polygon, on the inside of the clothing... on second thought, I wouldn't give the thighs any polygons at all. Make it all hip and just have the thigh joints there as ghost parts (bodyhandles).
Cinema4D Plugins (Home of Riptide, Riptide Pro, Undertow, Morph Mill, KyamaSlide and I/Ogre plugins) Poser products Freelance Modelling, Poser Rigging, UV-mapping work for hire.
who3d posted Fri, 11 November 2005 at 12:10 PM
Of course, if eFrontier could just fix the code that'd be another potential solution ;) Cliff
nerd posted Fri, 11 November 2005 at 12:14 PM Forum Moderator
Jessie's bikini probably isn't the best example because the mesh is pretty ragged.
The difference you are seeing between Poser 5 and Poser 6 with the Jessi Bikini Bottom is because Poser 6 allows you to limit the crease angle. Set the crease angle to 180 for the bits of Jessie's bikini and it looks just like Poser 5.
The problem with that mesh is that in the areas showing an artifact the mesh density is kind of weird. That facet that is getting burned because it's folded almost flat. The next facet over is not, it's about 70 degrees and therefore it gets smoothed.
So, if your mesh has places where the creases rapidly change sharpness just set the crease angle to 180 and it will behave just like poser 4.
Message edited on: 11/11/2005 12:28
Jim Burton posted Fri, 11 November 2005 at 12:15 PM
Colin posted Fri, 11 November 2005 at 12:21 PM
FWIW, I've seen the same thing in a model built in C4D 8.5, group-tagged in RipTide and imported into Poser for boning. Again, the problem ONLY shows up in Poser 6, NOT P5, NOT PP, NOT P4, and NOT D|S - only Poser 6 show this annoying artefact. For me, it occurred where one has a crease near (but in my case perpendicular to) a body-part boundary. The only work-around seemed to be to build in additional polys to fillet the crease, i.e. lessen the 'crease' of any individual adjoining pair. I'd be relieved to hear some statement from efrontier on why they changed this... (or more importantly, allowed it to remain) EDIT (after reading cross-posts): I just went in and changed the crease angle on the offending parts - by gosh, it works! Thanks folks!
Message edited on: 11/11/2005 12:35
nerd posted Fri, 11 November 2005 at 12:23 PM Forum Moderator
nerd posted Fri, 11 November 2005 at 12:30 PM Forum Moderator
Collin, just change the crease angle in the properties box for the object.
Colin posted Fri, 11 November 2005 at 12:36 PM
Nerd, yes, I tried it after reading your intervening posts - it worked. Thanks - much obliged!
Spanki posted Fri, 11 November 2005 at 12:43 PM
Glamorous Jessi has JCM in that area, so it is sorta critical where the joint is, it has to match the figure. Does that bikini have the same number of vertices, in the same order as the hip of GJ? If not, then you can't re-use the morphs from GJ directly in the bikini, so it shouldn't matter where the joint is (within reason) at all. Or even if there IS a (mesh) joint. Of course you'd still have to re-create any JCM in the bikini, but that's still doable. The mesh-splits don't define where or how the mesh bends (with the exception of how children affect thier parents). Only the joint-parameters and/or fall-off zones do. If you have "Secrets of Figure Creation with Poser 5" (great book!), what I'm talkng about is discussed starting on page 119 with a great illustration on pages 122 and 123.
Cinema4D Plugins (Home of Riptide, Riptide Pro, Undertow, Morph Mill, KyamaSlide and I/Ogre plugins) Poser products Freelance Modelling, Poser Rigging, UV-mapping work for hire.
nerd posted Fri, 11 November 2005 at 12:53 PM Forum Moderator
Here's what is happening. The sharp crease where the strings hook to the bikini bottom are being automatically divided into separate groups. The next polygon over is less than the 80 default and is not being divided, it's smoothed. This change from smoothed to unsmoothed produces an artifact. (One end of the facet is folded, the other isn't) When you see this happen don't look at the burned facet for the problem, look to the next facet over.
If you want, you can change the default crease angle in properties. You can also create smoothing groups inside Poser with the grouping tool. Using the smoothing groups is probably the best solution if your item is to be a Poser 6 only item. If it's to be a P4/5/6 item set the crease angle up so it matches the P4 behavior. Nerd3D
who3d posted Fri, 11 November 2005 at 12:53 PM
crosspost strikes again! Cliff
Message edited on: 11/11/2005 12:57
nerd posted Fri, 11 November 2005 at 12:57 PM Forum Moderator
who3d, see my post #36, we cross posted. I figured somebody would ask, "why" Nerd3D
who3d posted Fri, 11 November 2005 at 12:59 PM
My personal preferance would be for the default to be compatible with previous behaviour, personally :)
Cliff
Message edited on: 11/11/2005 13:02
nerd posted Fri, 11 November 2005 at 1:11 PM Forum Moderator
That's why you can set it in "Prefrences" and Once you save the object in Poser 6 the crease angle is saved with the object. Your customers don't need to change the crease angle in prefrences. Nerd3D
who3d posted Fri, 11 November 2005 at 1:18 PM
I've just had someone bashing Poser 5 & 6 for using compressed files "I'm glad I never upgraded from Poser 4!". My personal preferance for the DEFAULT would be for a product to be more compatible rather than less compatible with the previous version. That's why I said "default". Cheers, Cliff
nerd posted Fri, 11 November 2005 at 1:43 PM Forum Moderator
I think they chose 80 because it's more like the rest of the 3D world than the 180 that Poser 4 used. Max's default is 45.
And for those of you using MAX to create Poser content, try this. Select all the facets and set the Auto smooth value to 80 then hit the auto smooth button. That will give you a pretty good preview of what will happen in Poser when you import it. Jessie's bikini is burned in exactly the same places.
Message edited on: 11/11/2005 13:45
Jim Burton posted Sat, 12 November 2005 at 9:30 AM
Thanks, I'll try the crease angle setting! I still think of it as a bug, though- as Cliff said, the default should be compatable with previous versions of Poser. "The mesh-splits don't define where or how the mesh bends (with the exception of how children affect thier parents). Only the joint-parameters and/or fall-off zones do. Yep, exactly, but due to the way children bend their parents it does make a difference in this part. The way I do mtching JCM is it run the body morph to 1.0 (manually) and make a matching morph in the clothing, either with magnets or by exporting the figure to match it in Max. I then plug the clothing morph into the JCM channel which the clothing already has (as I start with the figure JCM) It is the only sure-fire way I've found, but the bending itself must be exactly the same.
Jim Burton posted Sat, 12 November 2005 at 9:43 AM
BTW, the crease angle setting doesn't seem to make any difference to me. However, I did install SR2, maybe this is now broke? I also don't see a folded polygon where the seams shows up, there is one nearby at the top, but not where the problem lies. Still say it is a bug!
who3d posted Sat, 12 November 2005 at 9:52 AM
ISTR I actually tried that setting on something which it now fixes at one stage and it didn't, but I probably only went up to 120 degrees or so rather than 180 and I've no idea which version of P6 it was :( I haven't tried it on anything this time around but I suspect it would be listed as a "feature" or "side-effect" rather than a bug per se, as it is perhaps "intentional" (and if they mean it to work that way then it's not a bug, however irritating). I note with interest that the way P6 SR2 responds to MAT files, and specifically to bump-maps applied via MAT files, has supposedly changed specifically to match the way content works in previous (pre-5) versions of Poser. This is exactly the way I believe it should have been set up in the first place. IMHO. Cheers, Cliff
Jim Burton posted Sat, 12 November 2005 at 9:59 AM
Funny thing is, on this whole deal, nobody seemed to notice it. Poser 6 has been out for a long time, after all. The reason I hadn't notice it was basicly because I wasn't using Poser 6 untill recently, to be honest. After I got Glamorous Jessi done (who was actually made in Poser 4) I did started to play with it, I find it very, very worthwile, BTW, dynamic clothing (which doesn't have this problem) is particullay facinating to me. But I noticed this problem on my very first item of P6 conforming clothing, and I quickly determined it only happens in Poser 6. I wonder why nobody else never seemed to notice it, Poser uses seem to go over figure textures with a microscope (by the number of super closeup renders I see), didn't they every notice the clothing was splitting at the edges? Now that Nerd pointed it out, it does seem to have something to do with how hard the edge is.
Jim Burton posted Sat, 12 November 2005 at 10:03 AM
Spanki posted Sat, 12 November 2005 at 11:00 AM
Yep, exactly, but due to the way children bend their parents it does make a difference in this part. What I meant by that was, a child can affect any/all of it's parent's mesh, but not any of it's grandparent's mesh. So a chest-twist (for example) can affect polygons in the abdomen (it's parent) but does not affect polygons in the hip (it's grandparent) at all, unless you create a JCM. This means that you might decide where the hip/abdomen split is, based more on how many rows of polygons you want the 'chest' to be able to affect. Similarly, a head-twist/bend can affect/deform all neck polygons, but none of the chest polygons, so that fact might (should) influence where you make the neck/chest split. In your example, there is ONLY a child/parent relationship (thighs and hip, respectively). This means that the thigh parts (whether they contain actual mesh or not) can affect any/all of the hip mesh, so it's not necessary to have any polygons at all in the thigh parts. For purposes of discussion, let's assume you use Poser magnets to create the JCMs. You would: - dial up the figure JCMs to 1.0 for the affected area (hip, thighs) - add a magnet(s) to the hip of your clothing (it's only body part that has actual mesh) - play with the magnet(s) until you got a matching morph - "Spawn Morph Target" on the hip - plug the clothing morph into the JCM channel which the clothing already has ...so it's basically exactly the same proceedure as you use right now, except that the morph only ends up on the hip, because there is no thigh mesh. You still need to have the thigh body parts there, set up with the right fall-off zones and other JPs, but they just don't have any matching mesh. When the (ghost) thighs are bent, they will still affect the hip mesh - EXACTLY as they would, if they contained any mesh. The only reason you'd need thigh mesh, is if these were biker (or longer) shorts, where the shin joints needed to deform the thigh mesh. (just a side note/reminder for others reading this... we're discussin Jessi/GJ, which don;t have any buttocks parts)
Cinema4D Plugins (Home of Riptide, Riptide Pro, Undertow, Morph Mill, KyamaSlide and I/Ogre plugins) Poser products Freelance Modelling, Poser Rigging, UV-mapping work for hire.
Jim Burton posted Sat, 12 November 2005 at 11:43 AM
Well, to explain myself a little better on this bending issue, when I set Glamorous Jessi I spent a lot of time setting up her thigh bending (which I happen to think is better than any existing Poser figure, incidently!). The outer bending zone actually extends into the mesh of the other thigh, but one thigh doesn't bend the other there (because it isn't a child), but IF it was all hip mesh it would, thus the bending would be slightly different, enough to throw off the fit in this bikini, which is rather critical, for the reason I showed. See what I'm talking about, now? Normally this would work, of course. Also, bear in mind, if I did do it all hip it would be in the nature of a kludge, don't forget. ;-)
Spanki posted Sat, 12 November 2005 at 12:17 PM
Ahh, ok... yes, if your fall-off zone overlaps the other thigh, then that would be a problem. But, isn't that also a problem in that, the hip (on GJ) is being totally affected by the left-thigh bends, and so the only real limit is where the right-thigh split is with the hip? (ie. doesn't the hip/right-thigh split end up with a sharp crease? or is the affect so minimal at that point of the zone that it doesn't matter?).
Cinema4D Plugins (Home of Riptide, Riptide Pro, Undertow, Morph Mill, KyamaSlide and I/Ogre plugins) Poser products Freelance Modelling, Poser Rigging, UV-mapping work for hire.
nerd posted Sat, 12 November 2005 at 3:30 PM Forum Moderator
Jim, not an autosmooth modifier, use the auto smooth button in the Mesh Edit, Polygon. (Waaaay down at the bottom) That actually creates smoothing groups automatically. It does a fair job of reproducing where Poser might burn facets. Nerd3D
Jim Burton posted Sat, 12 November 2005 at 4:32 PM
O.K. tried that, still don't get the effect on my mesh.
Jim Burton posted Sat, 12 November 2005 at 5:04 PM
O,K,, to get everybody on the same wavelenth: http://digitaldreams.bbay.com/test.zip Just a cylinder parted as clothing, with a proper edge formed. Incidently, the first time I did it I put a little bevel at the top and bottom (so the edges were about 75 degrees, not 90). Guess what, no bad seam. So that is a solution, but it is STILL a bug!
nerd posted Sat, 12 November 2005 at 10:09 PM Forum Moderator
OK, I see what you're looking at. In the Jessie bikini I think I was looking at the wrong artifact. For your tube example, set the crease angle to 180 and it looks exactly like P4/5. It seems like the auto crease thinge is getting confused at welded seams. The sharp edges need to be beveled or co-planered anyway to look right in P4/5.
For an edge like the one on the example tube, extrude the top facing row of polygons a tiny bit (Like 0.1 in max). That will make that edge co-planar. The tiny row of extra polygons take all the abuse and are too small to be seen in anything but a Sergio Leone flic.
This is really easy to do in Max. When you have your clothes flat modeled, select the whole thing and Extrude (along normals) for the thickness you want (1 or 2 max units) then invert the selection. (The edges you just edtruded are selected) If you want a diferent material the edges apply it now. Now extrude the width you want for the edge. Fianlly add the tiny "co-planar" on the edge. Edtruding one more time but a tiny amount. This will insure the edges look sharp in all versions of Poser. You can get really fancy and add more extrudes and bevels to make piped edges.
If you are interested in conserving polygons you can switch to edge mode and select open edges, delete, select open edges, delete untill all the back facing edges are gone.
Message edited on: 11/12/2005 22:10
Jim Burton posted Sun, 13 November 2005 at 9:18 AM
Yeah, I see now how to fix it, I mostly do a lot of that kinda edge stuff anyway, but I didn't on my bikini because it was going to be dynamic clothing at one point. But you agree it is a bug, right?
nerd posted Sun, 13 November 2005 at 1:01 PM Forum Moderator
Yeah. It looks like welded seams cause problems with the Automatic crease. It becomes really noticable if the edge is close to the crease value (A 90° edge and a 80° crease value). I can't imagine that it's supposed to do this, so it's a bug. I wish I'd know about this. I'd have pestered the programmers and maybe it would have gotten into SR2. You should probably send that cylinder thingie to EF. That's a pretty simple test case. If you'd like I can submit it to the Beta test bug report as well. Nerd3D
Jim Burton posted Mon, 14 November 2005 at 10:23 AM
Thanks, please do, link them to my file too (I'll leave it up). I had mentioned this problem to CL Steve quite awhile ago, but I guess it got overlooked.