Straps and Tanks and Annoyance, Oh My!

I understand that this is a failure on my part to fully grasp stuff related to the UV map, but I am having a devil of a time crafting shoulder straps on tank tops that look even remotely "right." So far, I've quite literally spent hours and hours on it, only to do an actual upload and discover that the results in-world pretty much stink, with a weird zig-zag occurring on the clavicle area of the avatar.

I'm sure that over time this will get easier, but I was rather hoping some of you with more experience could provide some suggestions, or perhaps even a very simple example template that demonstrates the right approach.

Thanks in advance for any insights.

You are running into an inadequacy in the clothing part's UV mapping. Getting thin shoulder straps to look good is impossible on certain parts of the shoulders, without doing strange distortions in your art to compensate for glitches in the UV Map there.

Try moving them out a bit or in a bit from the centerline. Sometimes that helps.

I have also had straps that look just fine... until I move. When I move in certain positions, the straps distort quite visibly. Move to a different position relative to the arms and shoulders, and it goes back to looking OK.

In short, you are not the problem. It's a limitation of the 3D Model.

I've grown the hate the UV map design. My pet peeve is around the upper legs, where there are a few 3-pixel wide UVs that stetch out over very large areas; without careful management the end results are horrible.

Learning to work with these militations is something you have to do; my advice is make a UV map skin with sub-divided UV, get on a posing stand, make a prim ruler, and then mark off the relevant bits in your file in photoshop (or whatever your graphics editor of choice is) to indicate where the ruler crosses them. then look at teh jagged line, shake you head sadly and continue work.

making a UV map skin is about the best advice you'll ever get as a beginner. its extremely helpful to see where things actually fall on the avatar in comparison to in your graphics program.

and don't get discouraged - i have yet to meet one designer in world who could say they managed to line things up properly on their first try, let alone 20 - 30 - 60. but, trust when i say it gets easier as you get more acquainted with the form.

btw - make sure you have the chip midnight templates as well - they changed everything lol.

http://www.slboutique.com/chipmidnight

Robin Wood gave me the following suggestion once:

Upload the detailed templates (either Robin's or Chip's) and wear them. Take screenshots. (My own personal suggestion would be to use high resolution screenshots.) Go into your editing program, and draw the outline of your desired clothing on on the screenshots. Then go to the actual templates and, using the drawn-on screenshots as a guide, draw the actual clothing.

The problem is that a) this is easier said than done and b) the clavicle area seems to create distortions no matter what you do. I don't know if it's possible to create a thin strap that looks good in all positions -- if you compensate for things to look normal with the avatar standing with its arms relaxed at its sides, for instance, things will still go out of whack when it puts its hands on its hips.

I have some experience with UV Mapping for 3D game assets (including one or two actually used in published games), and I think the quality of the SL avatars' mapping is horrendous.

I know that the avs are based on the figures that ship with Poser, and that might be the root of the problem. If I remember correctly, the mapping on the original figures that shipped with Poser is just as bad. At one point I thought that perhaps the mapping is distorted because it has to accomodate the scaling of the figure, but then I got a look at the wrist seam, where the sleeve lines don't match up -- that just looks like poor quality. My previous experience is with a community of game artists who would have nothing but contempt for the avatars' mapping as it exists now. Even with my meager skills, I could probably fix the UV distortions in a day or so. They bug me, because every time I see them I think "Bad UV mapping! FIX!" Assuming that they're not supposed to be that way in the first place, for some reason.

I could be wrong, and the UV distortion could be there for a reason. Can someone confirm or deny this? If it's not suppposed to be that way, then the Lindens need to consider how hard this makes creating quality textures for their avatars, and how much perceived quality it takes away from SL in general.

Okay, soapbox off...

It's been some time since I have played SWKOTOR2 and ES Morrowind, but I don't remember having the same sorts of problems using templates to create custom TGA outfits that I have had here, in terms of the weird stretching and mapping.

Maybe you could "volunteer" to fix the maps for them... for a nice tidy fee, of course.

The problem is really not the fixing of the UV maps, any decent RT3D artist could do that. I'm sure it's more a problem of implementing the change. Since the avatars have been a part of SL since its beginning, all the textures that have been created match those flawed UV maps, and improving the mapping would invalidate all the work that's already been done -- and marketed. Imagine what would happen if all humanity suddenly grew a third arm overnight, the apparel industry would go bust due to a surplus of obsolete inventory. Similar forces are in effect in SL.

My wish would be that Linden would allow new models that comply with the Poser-based requirements to be imported into SL. Not just anyone's half-assed 3D work, but something that could pass a Linden review process. You could have a whole new custom avatar, with better texturing. This would also fragment the market, of course, but it would definitely impact less of the membership. But then, I'm just blue-skying here.

Still, I hate the idea of having to deal with a lousy situation 'because that's the way it is'. But I guess it's just SL imitating life again, sucking in hard-to-reach places!

I'm glad I found this thread, now I know I am not insane. Or at least I am not the only one. I think I won't be a perfectionist about shoulders.

If they made a new and improved UV map, couldn't they add something so that different maps can be selected for clothing? That way existing clothes would continue to work (use "Clothing Fit 1.1", but new clothes could take advantage of improvements (use "Clothing Fit 1.9".

That's like saying "why don't they make new types of prim, and not let people build the old ones anymore, but keep the existing ones around for back compatability"

The only real problem I can see is mixing 1.1 and 1.9 clothes. But I'd take the initial pain to get a better avatar mesh with a UV map that was easier to work with and had two arms.

Yeah just looking at the wireframes in the Chip templates I can see the modeling is hideous. They need to hire a good game artist to model a nice avatar base. There are so many wasted plygons in the mesh, they could probably make the poly count lower and still have the meshes look better than they do now. Then having the model UV mapped with love would be wonderful.

Well... the Poser market already deals with this issue just fine. Look at any of the Poser type figures out there. DAZ Victoria 1.0 has a completely different UV map than the 2.0 version. The 2.0 version has many more polygons. The "Millennium" DAZ figures changed the UV Map again, upping the poly count even farther.

With each of these upgrades to the figures, the entire clothing inventory for the old figures was invalid for the new ones. Still worked for the old ones, just not for the new ones.

You know what happened?

People bought more clothes. People made utility apps that converted the old clothes UV maps to the new ones... They adapted.

Now in SL it's harder to do that, but not necesarily impossible.

Make a new avatar figure, with a higher poly count and better UV mapping. And both arms on the UV map! Get the new UV templates out to the designer community, so we can make updated releases of our clothes. Believe me, the designers may complain about the work of converting their stuff, but they will love the renewed sales!

The hardest factor will be enabling the SL Client to recognize that character "A" is a 1.0 version figure, with the old mapping, and character "B" is a 2.0 figure, with improved mapping. Can't be any worse than what DAZ Studio and Poser already dealt with.

Converting your avatar might require a complex task. Might be easier to allow people to create new accounts, and transfer the assets that are not UV-map constrained to the new ones. Ceera version 2.0 might need to get a new wardobe, and new skins and furry Avatars. Upgrading wouldn't be cheap. But by the goddess, she's be pretty enough to be worth the expense!

I't just life, or in this case second life he he. Stuff get's old an you have to replace it, which some times even means a redesign.

One thing to keep in mind is that the avatar mesh has to accomodate a huge number and range of morph targets. We're not just talking about a singular avatar shape and model. Every single shape slider in appearance represents an individual morph target (many of which have adjusted UV mapping to help compensate for stretching at the extreme end of the morph). While I've no doubt that the model and mapping could be improved you have to take into account the enormity of the job. All in all, despite some things that make me want to gouge my own eyes out, I think LL did a commendable job.

I've been thinking about it for a while, and I think that it would be possible to run both the "1.1" and "1.9" avatars in-world. I could be wrong, but we have both cubes and spheres, and both default and planar mapping. If the system can handle that, why wouldn't it be able to handle which avatar you are "wearing?"

Assuming it's a whole new AV (badly needed) as well as new UV maps, it would help ease the strain a lot if we were allowed to pick which one we use for a time. (Although I'm guessing that most people would leap at the chance to change, if the new ones were really the improvment that they could be.)

I would hope that the new model would be on the Preview grid for a while before it was adopted, with its model and map available for download. That way, designers could have new stuff ready to go when the new AV was rolled out; much the way that we all have a chance to "fix" our scriptable lights now, before lights all change in 1.9. (I don't know about everyone else, but my new candles are scripted, tested, and ready to put out the day 1.9 launches. I imagine I'm far from the only one who's ready.)

And by the way; I don't just want two arms. I want two feet, and two eyes, as well!

(Oh, and Poser maps were never this bad. I've painted a fair number of Poser 2 maps; they always had two arms, feet, and eyes, and although the head was on one map, the entire body was on the other, so there wasn't the problem we have with discontinuous seams over the jacket.)

Just my 2¢!

There's really no reason that I know of why it couldn't be done. In any 3d animation app worth its salt you can have an obect store more than one set of UVW's (I believe 3ds max allows 999 per object) and you can select which set any given texture obeys. They could just add a checkbox under the texture box in appearance (check for new mapping, uncheck for old mapping) I'm sure it's not at all trivial to implement but it's LL's best bet if they ever want to improve the avatar without making all the existing content obsolete. I suppose the other way to go would be to have two different avatar meshes and you chose which one to wear at any given time. I hope they wouldn't go that route since people wouldn't be able to mix and match old clothing with new.

I would love to have two arms, two feet, and two eyes, in theory, but would you trade resolution for it or is it worth the added bandwidth if the template no longer fit in a 512x512 area? We definitely wouldn't want to end up with a more convoluted template where it's all crammed into the same size. I think part of the reason the templates were done the way they were was to make it easier for novice texturers to comprehend, keeping all the body parts pretty close to their actual shape instead of going for accurate relative polygon size (for less smearing and smudging).

Everything is a tradeoff.