Distorted Sculpt

Okay, so I'll admit I'm new to sculpting, and blender is giving me absolute hell. I love mirror modifiers, but remapping the UV map after using one is nothing but a PITA.

But that's not why I'm here. So I've finished a sculpt, and it looks great in blender. I get it into SL and of course it just looks broken. I've even had a look at some of these precision tutorials, and snapping it to the grid simply makes no difference.

I upload with lossless compression too. any ideas what I could do to improve the look of it please?

http://www.users.on.net/~rye402/BlenderUV.jpg
http://www.users.on.net/~rye402/Door.jpg

Ah, Star Trek doors. Before you go too nuts with those, let me throw a little warning out. By sculpting them, you're spending 2048 polygons on each one. If you instead use a cube with a texture of a door on it, you'll only be using 108 polygons. The math is simple. It takes 19 cube doors to equal the performance hit of a single sculpted door.

If you're building a whole ship, you might have a couple hundred doors in it. Now you're talking over 400,000 polygons for the sculpts, vs. a measly 21,000 or so for the cubes. That's more than enough difference to make or break a decent frame rate for most of your visitors.

This is why you don't see many sculpted doors in SL. It's just not worth it. The detailing of a door doesn't need to have real depth in order for the door to be functional, or even for it to look convincing. Simulated depth via clever texturing can look just as good, and costs 95% less resource overhead. It's pretty much a no-brainer.

Also, don't forget that not everyone in SL is going to have their object detail slider all the way up. Those with lower settings might not see your sculpted doors display properly. But with cubes, any setting will work.

That having been said, if you really, really, really want to sculpt your doors (and I'll repeat one more time that it's not the wisest idea to do so) then what you've got there isn't a bad reproduction of the source model. It's not perfect, but it's not bad. I see some slight distortion in the decagonal recess in the middle, but other than that, it looks pretty clean.

I'm guessing your problem lies in the way you oriented your mesh when you started out. I can't be certain without seeing your mesh, but I suspect your orientation is 90 degrees from how I would have done it. If I were to sculpt that door (and believe me, I've certainly thought about doing it on more than one occasion), I would use spherical topology, and I'd plant the pole right in the center of that decagon. That way, the decagon itself gets to made up of a couple of edge loops. From there, the whole door is nothing but a series of concentric loops. The larger recess will be slightly tricky, but only slightly. The height of a few segments of each loop in that part will need to staggered a bit is all. And of course, the outer edges of the whole door are perfect looops.

The beauty of doing it this way is that since each loop is made of 32 segments, there's more than enough detail for the shapes you need. It won't last long against SL's LOD culling system, but since presumably doors like these would only go in relatively small rooms, that shouldn't be a problem.

I'd highly recommend you just throw a texture on a cube, and call it a day. Remember, success in SL is all about balance, about doing more with less. Don't get so caught up in trying to preserve every last mesh detail. There's not very much of that to go around. If you focus too much on the 3D geometry, you won't get very far. For best results, models should be kept as geometrically simple as possible, and texturing should be used to create the illusion of detail.

That's a bit of a leap isn't it? Not to mention very little to do with the question he asked and also you're thinking in terms of nurbs, not polys.

I don't think most people care how many polys something is if it looks good. Performance hit is the last thing on people's minds who strive for realism/detail with the fewest prims possible and most modern video cards don't even blink at high geometry. Huge textures are a far greater performance drain imho.

I've made plenty of scripted sculpty doors and the LOD doesn't pose a threat because I know how to get around it, and as you say, if the door is in a confined space, it won't be an issue performance-wise because of the occlusion culling.

So to the problem: You're trying to relief a plane or a cube by the looks, is it one sided or two? It looks like you have a little too much geometry on the flat areas so you should think about creating the door using one face per flat area with an oblong. You can shrink any unused vertices down to nothing (again negating any performance hit mentioned above) and make the door with around 120 faces total.

The mostly overlooked fact with sculpties is, that they have an intrinsic spatial resolution of 1/256 in each dimension. So if you choose the wrong orientation, your sculpties could get very rough and look imperfect. You have to consider this resolution effects when you design your sculptie. So starting with the optimal orientation is often crucial to the success of your sculptie work.

From what i can see in your case, you allready did a very good job. What you may try out to improve further:

- You could gain a lot of "realism" by using bump maps and baked shadows. So you could in principle make the rough polygon surfaces look more smooth by combining the sculptie mesh with an appropriate texture. You could get rid of many of the rough edges if you take care.

- If your door has got only a front but the back could be a simple flat surface, you could also consider to use a sculptie plane for the details on the front and back it with a simple flattened cube. Maybe put a texture there to 'emulate' the door shape. You would gain a significant amount of vertexes (but not necessarily better resolution) if you do so, so maybe in your case it might not be worth the effort to try that.

- If you need front end and back end both be perfect shaped, you still can go with plane sculpties and mirroring, so you end up with 2 sculpties one for the front, one mirrored for the back.

I personally second Chosen Few here. Is it really necessary to be so "realistic here ? Couldn't you achieve a simlar effect by using simple prims and appropriate textures ?

I have read somewhere, that the additional costs introduced by sculpted prims are not as high as expected. But after reading choosen's contribution here, i start fearing again that using sculpties comes along with huge costs regarding lag and performance... Is that true ?

When did anything about NURBS come in? Am I confusing my terms, or something? (Rather curious)

I took it from Chosen's post that he was thinking in terms of how a nurb based model would be constructed. you don't need to take much, if any of that into consideration with some well thought out poly modeling using the hardest LOD lines and so on.

Although reading back I can see why - the original model was flawed from the outset in some ways, but I really don't want to sound condescending or anything, there are some things best learned by yourself. I shouldn't really have got involved in the conversation as I don't really want to give all my secrets away!

No, I wasn't thinking in terms of NURBS. If I were, I wouldn't have been talking about edge loops, since NURBS surfaces don't have those. What I was thinking about was what appeared to me to be the most sensible way to structure the mesh, given the shape of the door, and the nature of sculpties. As I see it, that door is most efficiently described as a series of concentric rings, rather than an imprinted cube.

Does that description happen to have a lot in common with how the same door might be built from a NURBS surface? Absolutely. But that's only because sculpty topology and NURBS topology have a lot in common to begin with. In order to work well with sculpties, you need to think of each object as a flat rectangle, bent and folded in 3D space (because that's what it is). It happens that that's also part of the definition of a NURBS surface, but that's not why I suggested approaching it the way I did.

In any case, I must respectfully disagree with your assessment, Ponk, that just because modern video cards are good at crunching large amounts of polygons, high poly counts aren't a significant source of lag in SL. If that logic were applicable, the infamous Luskwood tree would never have caused all the trouble it did. If you're not familiar with what I'm talking about, it was a giant tree made from a few hundred pillow-shaped sculpties. It severely lagged everyone within a 2-sim radius whose draw distance was high enough to see it.

Changes to the LOD system have cut down on that kind of lag since then, but only over distance, of course. If you've got a few hundred sculpties in close proximity, your frame rate will suffer.

Keep in mind, it's not just the hardware that's at issue. Yes, modern video cards are capable of handling enormous amounts of polygons without breaking a sweat. But that only matters if the entire pipeline is equally capable, from software to hardware and back. In programs that have a lot to do besides just spitting out polygons (like SL), a high poly count will almost always cause trouble.

Think about the way a full featured 3D platform like Maya or Max responds to a very high poly count, vs. how a dedicated high-poly modeling program like Mudbox or Z-brush handle the same amount of geometry. In Mudbox (which I finally started using a few weeks ago, and absolutely love), I can crank up a mesh to several million polygons, and it just doesn't care. That kind of poly math is one of the only things it has to do, so it's free to do it extremely well. But if I try to put several million polys on screen in Maya, it will choke every time. Maya's got a lot more to do, so keeping track of all those polys isn't going to be as efficiently done.

Now think about video games. Low-poly modeling is still the prevailing law of the land in game assets, even though gamers have ridiculously powerful hardware. In fact, most of the advances in game art design techniques in recent years have centered around keeping poly counts as low as possible in-game. This is precisely why programs like Z-brush and Mudbox exist. High-poly detail models are created for the purpose of generating maps for use on low-poly game models. If games were capable of displaying those high poly counts at speed, none of these techniques would be necessary.

SL has to do all kinds of things that no other 3D modeling program does, and a lot of things that no game does. High poly counts do make it gag.

As for your point that people aren't predisposed to care about poly counts in SL, you're certainly correct. But that's pretty much my point. One of the biggest problems with SL's modeling paradigm is it often teaches new modelers to think about the wrong things. SL builders tend to think in terms of object count, which is completely arbitrary from a rendering standpoint, rather than poly count, which is universally crucial in every 3D engine in existence. With sculpties now on the scene, and other mesh forms on the way, it's super important to make every effort to educate people as well as possible to think about what they're doing from all applicable angles.

That kind of responsibility in thinking has already happened once before, at least partially, with textures. We all know that when people use textures that are too big, it's not good. So none of us here have any problem recommending that texture sizes be kept as small as possible. That wasn't always the case, though. It took several years, and a lot of deliberate effort (from myself and others) to pound that concept into the public consciousness. People still abuse textures, of course, but at least there's enough good information circulating now to keep it relatively in check. Builder culture in SL has evolved to the point where texture size abuse is considered a sign of "noobishness", for lack of a better term, which is how it should be.

There hasn't yet been enough time for a similar a similar evolution of "right thinking" to happen regarding poly counts in SL. Scultpties (and other eventual mesh forms that will be supported one day) are still relatively new in SL. So OF COURSE the average person doesn't yet think about the consequences of abusing them. It takes time for the right information to disseminate effectively. It will happen. We're just not quite there yet, culturally.

That's how I see it, anyway. You're of course free and welcome to disagree.

Chosen, you forgot to dedicate that post to family member or somebody.

And you forgot to explain what that's supposed to mean. Care to?

I wonder, if i could support some of the "right thinking" in a new tutorial, which i recently started to produce ... Maybe Chosen or other experienced builders have a good mind to add some input to the thread:

/109/b5/314984/1.html

I will cover some techniques for the creation of trees and maybe it makes sense to also cover "the right thinking about ressource usage" ? There are allready some very interesting informations available in this thread and i appreciate any further comment which helps to add significant knowledge to the tutorial... (of course every contributor will get credits on the video

regards,
Gaia

Wow.. There are certainly some strong points of view flying around about this. I certainly didn't intend to cause any form of arguments! Both lots of ideas, Chosen, and Ponk, I think are great for me, as I am still learning blender, and this kind of info is exactly what I need!

As for the poly count, I admit I was doing it this way for the whole realism and detail affect, and I believed it wouldn't cause so much performance trouble. I am, as you mentioned Chosen, very much an SL builder. I think pretty much in objects, object counts. For instance, the frame that you can see me still building in Blender in that first picture, is to be two prims. I am sculpting it because to make it out of prims, I'm looking at somewhere around 13? That is to have it shaped just the same with all the holes filled nicely. Same for a group of pipes, that I'm planning to make out of the one sculpt instead of 10 or so prims.

To be honest, half my reasoning behind making the door a sculpt was that Blender in my mind is so much easier than say Photoshop. Blender I can make my object shape, and have it do the texture along with the correct shadowing right there for me as well! But in Photoshop I can't make decent looking textures for the life of me sadly.

I think in relation to the door itself, as it is only the one prim either way, I will look into making a texture, to save on the polys. Though I do still need to look at things prim wise too, so the frame I would have to stick to the sculpts.

Really though, do you realise just how little info there is on this kind of thing? there's all these base tutorials to get you barely started. But there's nothing about going further with it. I mean, it took me three days to figure out how to remap my UV after using a mirror modifier. Even now I'm not sure I'm doing it quite right, as it take a -lot- of stuffing around to get the UV map right again.

But yea, thanks heaps for all the input! Trial and error is all I can do, and this is all a learning experience!

That is exactly why i asked Chosen and other experienced builders to comment on my other thread. I am planning for tutorials with much more significance compared to those we did last year. I could need a little help here

Have you been investigating on Domino's scripts and the techniques, he is supporting ? I wonder, if you can get something out of this tutorial:

http://blog.machinimatrix.org/2008/06/16/the-arch-example

The usage of the mirror modifier is shown starting from minute 12:15

Domino has created a small tutorial by himself too:

http://dominodesigns.info/downloads/tutorials/blender/head.avi

Or look at his thread

/8/60/203571/1.html

But i guess you allready know all of this ... no ?

I thought it was a slight at the length of the post... like how most books are dedicated to someone?

Maybe, but the post was all of 12 paragraphs. That's less than one printed page. It could hardly be described as lengthy. At average adult human reading rate for the English language, it should take about two minutes to read. Some people have underpowered attention spans, I guess.

Wandered, if that is indeed what you meant, it's unfortunate that you feel that way. Me, I'm not a part of the Twitter generation. I won't apologize for actually knowing how to write. Some things do take more than just 10 seconds to explain.

Generally agreed, but let's not forget that a hollow torus with four turns is >8000 triangles which is 4x as bad as a sculpty. So sculpties are not the worst polyhogs! Even a plain torus is 1024 triangles.

I built a five piece sash window for my home (the frame, two panels and two panes) using one sculpt mesh for the frame and one for the panels. It seemed to be a useful way of doing it because I couldn't get the right proportions using the usual building tricks of hollowed out boxes and so forth. Also, I have over fifty windows and the prim count using cubes would be far higher.

I haven't installed the windows yet for a variety of reasons but it would seem that this isn't one of my better ideas? Would I be adding fuel to the fire by also scripting the panels to slide up and down?

There aren't many visitors to the estate where I live but there are a few serious builders. In any case, I wouldn't want to be a source of lag for love nor money even if I was living here on my own Ben Gunn-stylee.

Chosen, I'm afraid the amount of words you type doesn't make you right and stating examples that are shining reasons why some people shouldn't be allowed near SL doesn't help your cause.

You could make the whole ship you mention from a single prim, if you like I'll show you. Until you fully understand what's possible and have a rounded grasp of every technique available, maybe telling people what they can and can't do due to irrelevancies is a bit daft.

I would like to see that. Would you show it also to me ? ...
ok,ok ... i mean "show it to US"

Wow, a lot to respond to here. I'll break this up into a few different posts, so as not to continue to anger the "your posts are too long" crowd. I think you'll see, though, that it's kind of silly to do it this way, since it ends up adding more and more unnecessary pages to the thread, but I'll certainly give it a whirl this one time.

I'll start with Ponk.




Who said it did? How could that possibly even make sense?

If you'd care to provide factual evidence to dispute any point I've raised, go right ahead. But arguing simply for or against any certain amount of words is more than a bit silly, don't yo think?



Where exactly did I say anyone shouldn't be allowed near SL?

Talk about "not helping your cause", making stuff up as you go along isn't exactly the way to get your point across, Pong. Again, if you'd care to dispute the facts of anything I said, please, go right ahead. If it turns out I was mistaken about any particular fact, I'll thank you for the correction. But just pretending I said things that don't make sense doesn't make it so.



OK, you go right ahead and make a ship with 200 doors in it out of a single prim. Give me a shout when it's done. I'll come take a look. I'd really like to see how you'll pull that off.



OK then, explain it to me. What's possible? What techniques don't I have a grasp of? Exactly which points among the ones I cited were "irrelevancies"? The floor's all yours. Make your case.

True (although my numbers from counting tris differ slightly from yours). But just because there's a potential "worse offender" out there doesn't make the sculpty's poly count any lower. Also, people don't tend to make entire builds out of toruses. (Well, except for hoochie hair.) Torus use tends to be sparse.

Luke, sorry I didn't respond to your reply earlier. It was late at night when I responded to that "dedication" thing, and I didn't see your post. I hope you didn't feel I was ignoring you. I certainly should have replied to you before anyone else. Here's my reply now:




Don't worry about it, Luke. Some people just like to argue for no good reason. Differing points of view are always useful, but some people unfortunately don't see it that way. They'd rather attack the person presenting an idea than discuss the various viewpoints. And as you can see, there is a very small but dedicated subset of them who like to freak from time about about posts that go beyond a certain length, which is really absurd. It's unfortunate.

So don't sweat it. It's got nothing to do with you.



Great! By all means, give both techniques a try, and see which one works better for you.



Here's what I normally suggest. From my observations, the "average prim" by usage has around 190 polys in it. Therefore, it takes 11 of them to equal the number of polygons in a single sculpt. So if a sculpty can replace 11 prims or more, it's worth using, from a poly count perspective.

That's not to say poly counts are the only things worth considering, though. Texture weight is equally important. And prim count certainly is certainly relevant as well, since SL is SL. It's all about finding the best balance for each project.



Here's one option you might want to consider, then. If creating textures in Photoshop isn't your cup of tea (yet), use Blender to do it for you. Here's how:

1. Construct your door in Blender any way you want. Don't worry about what it will take to make it work as a sculpty. Just make it the best looking door you possibly can.

2. Once your uber-fantastic kick-ass looking door is finished, throw in some good lighting, and do a really nice orthographic rendering.

3. Now, bring the rendering into Photoshop, do any post-processing you feel is needed, and then size the image appropriately to be a texture.

4. Upload the image to SL, and apply it to a cube.





The technique I described above would apply just as easily to the frame. It's something to consider.



You're absolutely right. If you're self-teaching, it can be tough to find good information on 3D in general, let alone on anything so odd-ball as sculpties. If you're trying to learn this stuff from scratch, it can be daunting. I'd say you're doing quite well, though, so far. You're certainly asking good questions, which is a sign that you're on the right track.

Blender doesn't have the best documentation, which doesn't help your situation. Gaia has done some really amazing work to start to change that, though. It's because of her videos that I no longer hate Blender. (I still prefer Maya, so I won't be switching any time soon, but at least now I know Blender's interface isn't so completely insane as I'd always thought.) If you haven't already, definitely go to Gaia's website, and go through her tutorials. They're REALLY good.




You're most welcome. I don't pretend to be a Blender expert by any stretch, but I'm always happy to help where I can. For what it's worth, you seem to have a great attitude about all this, so I have no doubt you're going to continue to do very, very well with it.

I'll be happy to help if I can, Gaia. I didn't reply to your other thread right away, because as you know, I'm hardly an expert on Blender. But if you feel my input could be valuable, I'll give it.

My advice, I'd say go ahead and install them since you already made them, and see how it impacts performance. 50 sculpts is about 100,000 polys. Depending on what else is in the scene, that might or might not be enough to kill someone's frame rate. Hopefully it's not.



Rendering-wise, my guess is it's probably not going to make much difference.

How are you planning on doing it, by the way? I take it the panel and the frame are both part of the same sculpty? If that's the case, then to switch between the two states, you'd either swap two sculpt maps, or possibly animate a single sculpt map? I'm curious what you've got in mind.




Benn Gunn style in SL, hmm. You'd need a herd of animated goats, at the very least, which would likely be a significant source of lag. I guess there's no way to win. Want some cheese?

OK, I think that covers everyone. My apologies if I left anyone out.

FWIW, I still think one long thread is better than five (now six) shorter ones.

Two sculpt maps are involved: one for the frame and one that is duplicated for the two panel frames. I was planning to use a sliding door script to make the two parts of each panel (frame and pane) moveable. It isn't absolutely necessary but mundane details like that appeal to me.

I also use a 1024 x 1024 texture for all the pieces but that includes two versions (one slightly weathered, one less so) which I plan to alternate here and there on different windows for a touch of variety. I ended up using a 1024 x 1024 because the vertices on the pieces are tripled to maintain LOD from some distance and even a 512 x 512 looks slightly blurred when it is spread over the visible quadrants of the sculpt meshes.

I was also planning to make sculpted doors and door frames. I think I delayed that along with installing the windows because some gut instinct told me it wasn't kosher. Reading over your comments above seems to confirm that.

Am I right to conclude that a lot of sculpts in one location isn't very good idea?