Chosen Few

Sassy Romano wrote: or run multiple avatar models in parallel. That's precisely what I've been suggesting, for the better part of the last decade. They can't just do away with the existing avatar, since as you rightly point out, to do so would be to render all content built for it useless. Also, from what I've been told, the existing avatar is deeply interwoven into the system, in ways it really shouldn't be, which make its removal nearly impossible, without a complete rewrite. So, if and when we get a better avatar, it would need to be a new selectable option, coexisting right along side the old one. Of course, if they implement custom skeletons, it almost becomes a moot point. LL could effectively get out of the avatar design business altogether, since users would theoretically be able to create complete avatar systems on their own. I suspect that that's probably the direction they'll go. All that said, if it were up to me, I would just declare SL as we know it "complete", stop work on it, and start a brand new "SL 2.0" grid, with completely modern technology, from top to bottom. No doubt a lot of people are glad it's NOT up to me. However, I do believe that's going to have to happen, eventually. There's only so long you can keep patching a leaky boat before you just need a new boat, and that's that. So much potential functionality in SL is held back by the lingering results of shortsighted architectureal decisions that were made over a decade ago. It sure would be nice to be able to break free of all that.

Nalates Urriah wrote: The mesh layout of polygons has problems and could have been modeled better, by today's standards. It was more than adequate at the time of creation. Hi Nat. Sorry, but I have to take issue with the above two sentences. The avatar model was never what I'd have called "adequate". It was poorly made in comparison with the standards of 10 years ago, and it remains just as poorly made today. At the time it was created, there were any number of better-made human models available, which LL could have purchased, or they simply could have hired a more competent artist to make one from scratch. As the story has been told to me by certain former Lindens, who shall remain nameless, the avatar was made by another former Linden, who shall also remain nameless here. Long story short, the guy bit off more than he could chew when he took up the task of creating a morphable human model. He'd overstated his abilities, and he ended up delivering a very sloppy product as a direct result. Before that, avatars in SL (then called Linden World) were made from prims. Stack a few spheres and cylinders together, snowman style, and you've got something you can represent yourself with in-world. It's not hard to imagine how thrilled the Lindens must have been when one of their own said something to the effect of, "Hey guys, I can make it look like a person." They'ed been staring at nothing but prims for months or years at that point, so it must have seemed like a godsend in comparison. The question of whether or not it was actually a good quality polygonal model probably never even orrcurred to anyone untul much, much later. Remember, this was back in the days when individual Lindens were allowed to work on whatever they wanted. Had that particular person not taken it upon himself at that particular time to create an avatar model, it's very possible SL might have launched without one, and we'd all have been hopping around as snowmen, or bouncy balls, or rocket ships. In that context, I suppose we can be grateful for at least the subpar human model that we did get. But subpar it was, nonetheless (and still is). The fact that we're still stuck with it, ten years later, is of course ridiculous. LL, if you're reading this, my offer from years ago is still on the table. I'd be more than happy to create a really nice new avatar model for SL. I've done plenty for other platforms. You guys know how to get hold of me.

Add some extra invisible geometry to the model, so that the whole becomes larger than .01M. All you should need at a minimum is a couple of perpendicular triangles. I'm not able to test at the moment, but in theory, SL shouldn't care if individual polygons are smaller a centimeter; it should only care about the size of the model as a whole. So, in Maya, just add whatever extra geometry you need in order to make the whole thing larger than a centimeter on each axis, and you should be all set.

MyAlt4099 wrote: I exported my avatar as a .obj using a tutorial I found online. Is there any way I can rig that automatically without having to mess with weight painting? Grab the standard sizing kit from the marketplace. Among other things, it will include the default avatar, fully rigged, in FBX format. Copy weights from that to the avatar model from your OBJ, and you should be good to go. Of course, any time you copy weights, you might need to do a little ceanup work by hand afterward, so do keep that in mind. Depending on what software you're using, which particular copy method you use, and the geometry and UV layouts of the models, copies can't and won't always be perfect. A little bit of repainting here and there will very often be necessary. With that in mind, if your goal in asking the question was to avoid learning basic begginner level rigging techniques, such as weight painting, you were barking up the wrong tree. There's no way to get around the fundamentals. You might as well be trying to bake a cake without first learning how to use a measuring cup. It's that crucial. As with so many things in the digital arts (and in life in general), one must either be willing to make the commitment to spend a few very short weeks learning the fundamental basics, or else just decide not to try to get into it all. There can be no in-between in this. You either do it, or you don't. If, on the other hand, you already know the basics, and you were merely asking for a tip on speeding things up a little, great. You, of course, can disregard the previous two paragraphs, and consider it advice aimed at anyone else reading who might have been thinking there could be any such thing as shortcuts in this stuff. We both know there are not. MyAlt4099 wrote: Barring that, is there anyone who, if I email you a .obj, you could rig it for me and send me the .blend file, either for a favor or for a small fee? This is the wrong forum for that. If you're looking to hire someone, head on over to Wanted or Inworld Employment. That's what they're there for. The Creation forums here are strictly for education. So, if you do want help learning the how-to's, then by all means, ask here.

I'm wondering, does mesh poo smell worse, better, or the same as sculpty poo? And prim poo? Or does nobody poop prims anymore?

Multiple heads is really your only option, at this point. SL does not support morph targets (shape keys) or custom skeletons. I wouldn't be worried about what happens when people turn on Highlight Transparency (ctrl-alt-T). It doesn't look "strange". It looks like what every model that has transparent parts looks like. The only reason people turn that feature on is when they want to see what's going on under the hood with a model, transparency-wise. Nobody has it on permanently, at least no one in their right mind. What you should be worried about is lag. Added transparency means added render passes, which tranlsates to lower frame rates in the viewer. If the model has a high poly count, then the problem only gets amplified. Ask yourself if it's really worth sacrificing performance, just to give a character model a few expressions. If you firmly believe it is, then go for it. But when in doubt, it's generally better to err on the side of performance, especially when you consider that you are not the only one who will be affected by it.

Screenshots would be helpful, so we can see what we're talking about. In the mean time, my guess is you've got overlapping UV's. You said you UV'ed each part separately, before you combined the meshes, right? Well, unless you mapped each one to a separate portion of the canvas, and/or unless you gave each one a unique material, all the maps would naturally overlap each other. If you didn't do either of those things, do them now, and the problem should go away. As for planar mapping, don't use it unless the model itself is also planar. A side of a cube, for example, is a plane, so planar mapping can work on it. A sphere or an elephant or a gorilla is not a plane, so planar mapping cannot work properly on it. Planar mapping in SL is just like when you do a planar projection in Maya. If my theory about your overlapping UV's is correct, then the reason planar mapping appeared to make the problem slightly better is just because the projection, by definition, ignores UV's

Thanks for following up, Chic. To answer your question, Maya does indeed assign a 50% gray Lambert shader to surfaces by default. If you upload a model that has just that default shader on it, and you try to include textures with the upload, the result in SL will be a model with a blank texture, with a 50% gray tint. If he wants to apply his AO maps as textures, he'll need to upload those separately, and then apply them, just like any other textures. This, of course, requires that his UV layouts and material assignments are set up properly to receive the textures, in the desired configuration. If he wants the AO maps to be already on the uploaded model, as soon as it appears, he'll need to apply them to the approprioate surface shaders in Maya first, so that the .dae file will contain the correct references to them. If he's using Turtle, this is super easy, as Turtle has a setting to automatically apply baked textures to the model, as soon as the bake finishes rendering. If he's using any other renderer, then he'll just have to do the assignments by hand (which, needless to say, is fairly trivial). Otherwise, he can just upload the textures separately, and apply them in-world by hand, which is generally the recommended approach, anyway. All that said, I'm still not quite sure how the presence or absence of a roof is factoring into any of this. If the roof puts the model at too many materials, or too many vertices, for SL, then he should simply break the model into pieces (or better yet, remodel it so it doesn't need so many resources). If the roof is screwing up his AO bake, then as I said to Fawmusi, it's likely that either his AO settings are wrong for the size of the model, or he's including more than just AO in the render output. If it's something else, which we don't yet have enough information to diagnose (which I suspect is the case), then of course, we'll need further explanation. I'd suggest you invite your friend to post here, and throroghly describe what's going on. Let him know there's never any shame in asking for help. I know it can be a matter of pride for some people to find their own answers, but really, if it's a difference of solving a problem in minutes or hours instead of days or weeks, then it becomes shameful not to ask. Besides, if he really feels strongly that asking for help would be embarrassing, he can always use an alt, and none of us ever need know who he is. If none of the above is acceptable or helpful for him, then as you said, he's just going to have to work it out on his own.

Frawmusl wrote: AO he won't get anything due to the roof, of course. The roof is covering the entire interior. I'm not sure why you'd say that, Fawmust. AO measures how much light can escape from a surface, not how much light is actually hitting it. Where surfaces are close together, such as at corners, less light can get out, so those areas darken. Where surfaces are further apart, such as the middles of walls, more light can escape, so those areas brighten. This is equally true whether the surfaces in question happen to be horizontal, vertcal, or otherwise. The mere fact that a roof will block external directional light from hitting a floor doesn't necessarily mean it will cast AO onto the floor. It depends how close the two are together, and on the AO settings you're using. If your experience has been that roofs/ceilings always exceessively darken your AO bakes, then I'd suggest either your occlusion settings are not right for the sizes of the models you're baking, or perhaps you're including something other than just AO in your render output.

Chic Aeon wrote: When he uploads the house with the roof (I am assuming joined) he cannot get a bake on the interior walls, ambient or full. I'm afraid I don't quite understand your wording here, Chic. Was he unable to create the bake in Maya, or did he create it, and simply cannot apply it in SL? I can think of a few things that could cause either scenario. Any more information you can provide would help narrow things down. So you know, I might not be able to respond with my usual level of detail. I was in an accident, and my right arm is rather uncooperative at the moment. Typing is painful, so I'll have to keep it brief. Happy to do what I can, though. The more info you can get from your friend, the easier it will be.

Subtracting is almost never a good idea when weight painting. It can be unpredictable, and unreliable. It's nearly always better to do the opposite. Instead of subtracting weight of wrong bones, add weight of right ones. Make sure your brush is set to replacement mode, or whatever your particular modeling program happens to call its equivalent setting.

Pagemaker? That's going back. You won't be able to open the document directly in SL. I'd suggest you export each page as an image file, and then upload them as textures. Another option is to export to html, publish to the web, and use html on a prim for display.

In addition to what Code said, you could try smoothing your meshes, to effectively create a high-res version for texture baking. As long as the UV layouts are preserved, the textures will fit the original low-res models. Also, if you have Turtle, you'll find it's a lot frieldlier for baking than Mental Ray is. Turtle is the only renderer that was built specifically for baking, from the gound up. If you don't have it, though, then never mind, of course. You can still get great results with MR. It just takes a bit longer.

Ah, thanks for the correction. I didn't know that.

Scukpties do not have customizable pyisics. Every sculpty has the same physics shape, which is equivalent to that of a torus with zero hole size, no matter what the visual shape might look like. The visual shape of a scpulpty is just a rendering effect. The shape is not "real". If you want a surface that can be walked on, use mesh, not sculpties. You can make the physics shape of a mesh model be anything you want. This is just one of the many, many ways plain old mesh models are vastly superior to sculpties.

If you're talking about the one I'd been hosting for several years, I'm afraid my site got nuked when I forgot to pay my hosting bill, and I haven't had time to rebuild yet. I've put a few Maya <-> SL things in my Dropbox, though. Here's the folder link: https://www.dropbox.com/sh/6tt1k8vwlbkof0u/nit7l93DJc You can also grab the Standad Sizes kit from the marketplace, which includes the skeleton, and a lot more. It's got MB and FBX files in it. If the MB's don't work with your version of Maya, the FBX's likely will.

Good catch, Ivanova. That may actually be my fault. Spinell followed the advice in my post, and I neglected to mention that the two objects should be combined into one mesh. I've edited the post, to include that step.

Innula Zenovka wrote: I've just been doing a bit more poking round reseach and come across this free plugin for the free programme, Paint.NET. It's almost as easy to use as the photoshop plug-in, and far easier than any of the manual techniques I've tried over the years (I think Paint.NET is Windows only, though). I hope it can work better than it did for their "final result" example photo. The people in that pic look VERY roughly cut out. FIne hairs, frayed edges, and subtleties in perimeter shading all got nuked. If I were to submit photos in that condition to my clients, I'd get fired. I hope it's just a badly done example, rather than indication of the inherent weakness in the plugin itselt.

Nice find, Innula. Thanks for the heads up. I'm looking forward to giving this a whirl. It looks like a great time saver, if it works as well as the demo video indicates.

...And suddenly I'm hungry for deviled eggs. Curse you, Madeliefste, and your little eggs, too! Anyway, they look great. Personally, I'd question the value of investing polygons into those little leaves, when an alpha texture on a flat plane would likely look just as good, but I know not everyone in SL agrees with me on such things. Regardless, nicely done.

Madeliefste Oh wrote: I'm not familiar with Illustrator at all, and I don't have to touch it with the method I'm using now. I just have to save it in PS as .ai file, and then can import it right away. Gotcha. I wouldn't suggest learning Illustrator, just for this, of course. However, if you do want to start using it, you'll find it's really useful, and the knowledge you already have from Photoshop's work paths will directly apply. It's kind of like work paths on steroids.

As others have stated, the most obvious issue is that the roof of the house is a nonmanifold N-gon, which is impossible to triangulate. That's not the only issue, though. You should make the following easy adjustments: 1. Detach the floor, and make it a separate object, so that you can reverse its normals, without creating non-manifold gemoetry. You need the normals to point upward. 2. Add four edges to the floor, to create five quads instead of one giant N-gon. Do the same for the roof. 3. The vertical edge on the wall near the right front corner creates an N-gon on the roof above. That ednge serves no purpose, so get rid of it. 4. The vertical edges above the front are also creating N-gons in the roof. Move their top vertices to the nearest corner, and merge them with the conrer vertices. 5. Once everything else is done, combine the floor with the rest of the house, so it's all one mesh. Do NOT merge any vertices at this point, though. For best results in-world, you'll want it all to be a single mesh object, but in order to keep the geometry manifold, the floor and the walls should not share any common edges, since their normals are pointing in opposite directions. See the animated GIF image below. The animation toggles back and forth between how you have it, and how it should be. You're also going to need physics for the interior sides of the walls. Edited (several times) to include the fixing of numerous extraneous edges, which had escaped my notice, the first time. ETA: Step 5, which I somehow forgot to add earlier.

This has me curious, as well, what could trigger a triangulation error. I wonder if funky ngons or nonmanifold geometry could do it. Must experiment when I have some time.

You'll find it works extremely well for trees. I've used it literally thousands of times for that, always with great success, very quickly. A couple of years ago, I took a full time gig as director of marketing for a real estate company, for about six months. With few exceptions, real estate agents tend to take really crappy photographs, so I spent the better part of each day making dozens of bad photos look like a million bucks. One of the things I constantly had to do was turn gray skies blue. To do that, I'd quickly channel-mask the skyline and foreground elements, which were nearly always chock full of trees, and then I could make the sky any hue I wanted, without affecting anything else. Selecting all those trees any other way would have taken hours per image, instead of seconds or minutes. Ditto for things like chain link fences, telephone lines, roof antanneas, you name it. Anyway, I'm glad the info has been helpful. Let us know how you make out.

Definitely give it a go, Trinity. Just keep in mind, it works most easily when the subject and the background have differing colors. If the background and the subject have very similar coloring, it may take a bit of finessing to arrive at the selection you want. Make sure to take a careful look at each channel. It's possible that a pair of colors may be identical, or nearly idendical, in one or two channels, but quite different in another. For example, take a gander at the image below. If your monitor is of average quality, the two reds probably look the same on your screen. They are 100% identical in two of the three color channels. It's not until we examine all three channels that we see there is actually quite a bit of difference. While both colors have the same amount of red and blue in them (100% and 0%, respectively), they have very different amounts of green. One has zero green, and the other has about 25%. That difference is more than enough to let you know where your seection should be. Of course, when it's just two large blotches of color like this, any number of selection methods could work just as easily, but that's not really the point. In a photographic image, it won't ever be this simplistic. In more complex imagery, diving into the channels is often the only way of fomring a really good selection, at speed. Also, keep in mind that you don't have to limit the process to just one channel. Depending on what's going on the image, sometimes you'll want to use one channel to grab one part of the subject, and another channel to grab a different part. To combine the results, simply shift-ctrl-click the thumbnails of all the alpha channels, and create your mask from the whole kit & kaboodle. Once you get in the habit of working this way, you'll find it's very powerful.