Noob modeling/texturing thread

[Baddcog]

The reason behind me saying you should use an alpha plane for details is that I imagined the details would be paint and/or goldleaf (or silverleaf) designs. In which case you really don't need normals (bumps) for the details.

Also because the wood on a lute would be sanded very smooth so you wouldn't get woodgrain bumps thru the paint.

It will also allow you to use an existing wood texture and ONLY make a texture for the details (256x256 is probably fine) If you painted the entire lute on a tex you'd probably need to make a 512x512. So it'll save space this way also. The 2 polys MIGHT add 2kb to the model, a 512 map would add at least 1 MB over a 256.

 

I made a wooden sign which I wouldn't do this on, because it is made from rough boards. I want the paint to follow the shape of the cracks ect... So it needs to be applied to a wood texture that matches the normal map.

 

 

@Max

For the sword I'd probably NOT add planes for details. Instead you probably want to use a plain metal diffues (no details) and any shapes in the sword will be in the normal map. That'll make it look forged 'cause all the details show in the shadows.

The specualr only makes parts shiny or not. So a leather handle specular would be black (not shiny), the polished metal blade would be white. A rusty blade would be specaled grey/white/black

 

 

Well, in 3dsMax it is the material browser, a pop-up window that has slots for each texture. Each slots material has properties, one of them (the important one for games) is the map. This is where you load your texture (you can drag from a folder and drop into max). Then drag it onto the object.

Then use the uvw tools to align it on the object.

 

@demagogue,

Since you are using LW you have NO choice but to triangulate if you want to import into D3. (3dsMax files work fine without this step).

 

I'm not sure how LW works in this aspect. But in 3dsMax I can turn the edges in the center of a square that aren't outlined (they can be guessed as to their location, or seen from the lighting sometimes). But it is tougher, so triangulating them (turning into 3 sided polys in Max terms) just makes them easier to work with, that's all really. But sometimes it makes detailed models harder to work with cause there are so many lines on screen (that's the main reason to work with higher sided polys).

 

A little unclear about what exactly you mean in that last part, hope I'm getting it right:

The engine DOES care about how many polys you have, thus the need to be conservative and try to build as low poly as possible while still getting the desired shape and lighting props.

Of course times are changing. In Max's 'something for ladies' post I posted a pic of a chest I made. The finish one for game is about 800 polys, maybe a little too high but it's a fancy chest that I hope won't end up in EVERY room of a castle. The high poly version for the normal map is 25,000 polys. It loaded in game just fine and could be used in a small area just fine probably, but overkill.

 

I think I get what you're asking after reading again. The engine will render the light and shape based on how the triangle of the back of lute are. No matter if the are triangulated or not. They are there, just hard to see without being outlined. So I'd probably finish fleching out the basic shape and get it right where you want it. When you are done shaping, then triangulate and turn the edges.

Always turn edges that need it.

---------------------

This is why GOOD low-poly modeling is usually thought to be harder than high poly. With high poly you can add all the curves, shape, ect and make something look real. Low poly really makes you work to get a convincible and good looking object, textureing also plays a huge part because it can mask the fact that something is low-poly.

This is a really interesting thread you guys might like, Low Poly Models : Polycount forums

http://boards.polycount.net/showflat.php?C...;gonew=1#UNREAD

You can see lots of great work there, along with pics of the maps and wire frames to get an idea what those guys are doing. This is one pic I thought was cool, old school low poly

 

 

Oh yeah, @Max

you probably only wnat to load a normal in the editor. Like in the pic of the doors, I used the normal for the hinge because that's where all the shape is, I wanted to align that. The diffuse is plain metal so it doesn't show you any reference.

 

Then you need to make a material file, combining all your maps (diffuse, normal, spec) into a material. The name of the material is what you put on the object before export (in LW I believe - I know its different than 3dsMax, for 3dsmax I have to edit the object files with notepad and put the material names in myself). I think LW can handle the material names somehow, it is on the Wiki.

[Springheel]

So it'll save space this way also. The 2 polys MIGHT add 2kb to the model, a 512 map would add at least 1 MB over a 256.

 

Saving space is not a bad thing, but reducing polys is more important because it limits processing requirements while the game is running, as you know.

 

You'd probably get decent results from a 256 x 256 diffuse for a model like this, especially if you mirror the uvmap. That's one of the best ways to save space when you're making symmetrical objects.

 

The name of the material is what you put on the object before export (in LW I believe - I know its different than 3dsMax, for 3dsmax I have to edit the object files with notepad and put the material names in myself). I think LW can handle the material names somehow, it is on the Wiki.

 

Your material has to be called the same name as your surface name in LW. You can have as many surfaces as you want, though multiple surfaces can't point to the same material (you'd have to combine them into one surface).

[Maximius]

Thanks for your responses guys. Im not certain but I suspect that I am going to have to install D3 again to actually use these mtr. files, is that correct? Forgive me if thats obvious. This means I'm going to go ahead and do a complete OS re installation, the puter has been a bit slow recently anyway and its heavily cluttered with crap so its a good time for a number of reasons.

 

 

cog, bear with me a second longer. So only normals can be loaded in LW itself (assuming .mtr's are for D3) So if I wanted to really pull out the pitted detail of the cast iron, I make the texture at least 512 by 512, which would mean I have to tile it, then make a normal map of it, then load both of these into the image editor.

[Maximius]

I did it, I got the cast iron texture into LW and the normal map too. Just as you insert the image file into the color T box submenu image, for the normal map you do the same thing under bump T box submenu, all of this is on the surface editor of course. Theres a bit of a difference but its kind of hard to see, the texture I made is the first image, the second is LWs stock.

 

http://aycu20.webshots.com/image/23259/200...90801298_rs.jpg

 

http://aycu17.webshots.com/image/24936/200...73705582_rs.jpg

 

Im guessing that a specular map could be loaded under specularity T submenu image too, Im going to make one of those next. This if fun!

Edited August 4, 2007 by Maximius

[Baddcog]

Well, you can load all the images in the 3d program. I can load 10 different ones in 3dMax including illuminations, ect... But I don't think they will import correctly. That's all in the mtr file for the game, not the model.

 

In fact if I do load all the maps in Max, I can see them whn rendered but only diffuse when working, that's why I suggest using the normal IF it has details that need aligned. Obviously for bumps only it's not that critical.

But if I load all the maps it subdivides ,materials in the object file on export and I have to clean up alot of stuff.

(easiest way is to reimport i a new file, reload ONLY materials I need, then re-export clean)

 

Congrats though, always fun getting something new down eh? Yeah, spec is same way.

 

Its true as I stated above in my low-poly rant, and as Springheel said that using fewer polys is best. Although I hardly think 2 polys is gonna make a difference with an engine that can handle models up to 25,000 polys. If I even tried to load a model in T2 with 2,500 polys it would just crash. The important thing is just to try and keep em all fairly low so when you have a bunch of stuff on screenit doesn't turn into 1000,0000 polys ya know.

[Maximius]

I see what you mean, but its still great to finally see this stuff coming together. Give me a few days to reorganize my files, then its OS replacement time, then I'll reinstall everything including D3 and start making some mtr files.

[demagogue]

Im guessing that a specular map could be loaded under specularity T submenu image too, Im going to make one of those next. This if fun!

 

You got it. I've been having the same fun the last 2 days just loading in different images for the different maps and seeing what I can cook up.

 

For the bump map, I've been using a gray-scale height map (white is high, black is low, shades of gray are gradients in between) and that seems to work so I can see it in action and align things (which is important for me). I don't think specularity is good for the wood I'm using though, just the metal frets.

 

edit: Triangulating just popped up my poly number from 480 to 940.

Looks like I have some work to do. I can already see places to cut 100s at a time.

 

Edit2, much work later: Ok, back down to 422. Don't ask me how!

Edited August 5, 2007 by demagogue

[Baddcog]

triangulating shouldn't increase polys. Just make them all visible. Not sure what you've got going there.

 

Yeah, spec on wood is generally bad.

 

I have mixed feelings about normal maps vers bump maps.

 

Normal maps do seem to work pretty good, and it's cool that you can bake normals from a high poly obj onto a low poly vers.

They are harder to create manually in Paint progs because you are using a wide variation of colors to get depths and correct lighting left to right and top to bottom.

Normals seem to be the wave of the future though.

 

Bump maps are cool because it IS easier to paint what you want in greyscale. You can also do the white against black trick to make a 90 degree edge. Normals CAN'T do a 90 degree edge. They can come close, but the edge has to be at least one pixel wide to work because the high and low surfaces are the same color of blue. blue against blue shows no shadow.

 

AFAIK, D3 does allow normals and bumps to be combined in one tex. While it might lead to very good results it is alot of extra work and the benefit would probably be lost on the player anyway. Plus the extra file size...

 

I have released objects that use bumps and some that use normals. I used bumps on objects that were already complete for the most part. It was easier than building a high poly around the existing low.

I have used normals on most newer stuff.

Either way I think the results are similar in D3 and it doesn't really matter too much. File size seems about the same too.

-------------

I mentioned earlier that spec maps were greyscale, but in fact they can be fully colored. Didn't mean to be misleading. If you have a red jewel on your sword you could give the specular red where the jewel is and it will have a reddish shine.

I have a few models that I should probably tweak in this way to give em a little more life.

[demagogue]

triangulating shouldn't increase polys. Just make them all visible. Not sure what you've got going there.

 

You are basically right (maybe one little footnote I put below). It's just the way LW counts polys for statistics purposes, what's visible as a poly, and I was just writing that number down because it was right at hand.

 

.......................................

 

I'm learning this as I go.

Here's how I understand it so far.

I'll try to explain it to see if I have it right (this is more for my own sake; no questions here so no need to read it if you don't want to).

 

The way LW works, "polygon" is more of a labeling convention than anything about the geometry per se. You could distinguish planes (of multiple polys), planar-polys, and non-planar-polys.

 

A "poly" itself, the thing LW is counting when you hit statistics, seems to be just an internal or nomenclature thing special to how the program builds the geometry (not specifically what it results in). You build surfaces by linking a set of points (*any* set of points) together, and it draws a surface and then it calls that surface one "poly" ... doesn't matter if it's not planar, if it's contorted into some ungodly shape, a "poly" can be a single line, it can even be a single point I've discovered (although these latter two polys won't render, you can give them properties to do other things, and rendering a non-planar poly, it tries to make it look like a plane, but rotating it around you quickly notice it's just not right). The only thing that matters is that you highlight some points and press "make polygon" (or however else you make them), and it calls whatever surface results a polygon, and counts it towards the statistic-count.

 

So at least the way LW counts for statistics purposes, when you divide one square into two triangles, it counts it as "1 poly" changing into "2 polys". That's what I meant in saying that my triangulation doubled my poly count. That's just how the program counts what it labels "polys" for its own internal purposes, even though the geometry itself isn't changing. It's just its own labeling convention. Ostensibly, it had been that doubled number all the while.

 

What doesn't change (usually, anyway) is the plane. This is what I'm catching on to. All triangles are planar, by definition (3 points=plane). But it is possible to make surfaces in LW that it calls "polys" with 4+ points, some will be on the same plane, but others don't have to be. So when you triangulate them, the planar polys turn into triangles that are on the same plane as before, but the non-planar polys turn into triangles that are now on different planes. This is the one little footnote to what you said, because that might be more like making new polys because they're on new planes.

 

What ultimately matters, I'm discovering, are the verticies, the points that define the plane the poly is on. This explains a confusion I had above. I naively thought that turning all those squares on the back of my lute was going to give me more polys to work with to make a smoother surface, just under the bad analogy that if you could get closer to a curve with more lines, you could get closer to a sphere with more surfaces. But actually, after I triangulated all those squares, it's just visually obvious that, even with LW telling me I have 2x the number of polys, the verticies defining the planes are exactly the same, and so I really don't have any more surfaces to work with. Bad logic. The most I could do is rearrange the space I have with those triangles to be more even, but I can't get closer to the curve than I already am.

 

Anyway, the point is I know what you're talking about now.

Edited August 5, 2007 by demagogue

[Springheel]

AFAIK, D3 does allow normals and bumps to be combined in one tex. While it might lead to very good results it is alot of extra work and the benefit would probably be lost on the player anyway.

 

This has been used to good effect with some of our AI models, like the revenant. However, I'm not sure it's any different than turning the heightmap into a normal and using the addnormals command (or just baking them together). From what I've read, D3 automatically converts heightmaps into normalmaps anyway.

 

Bump maps are cool because it IS easier to paint what you want in greyscale.

 

You can always paint your greyscale heightmap and then run it through the nvidia filter in PS. That's how I created a lot of my hand-made normalmaps.

[Maximius]

You can always paint your greyscale heightmap and then run it through the nvidia filter in PS. That's how I created a lot of my hand-made normalmaps.

 

Is this possible through GIMP?

[Springheel]

Not sure. There's probably some kind of normal-map plugin though.

[Baddcog]

 

having some IE difficulties here... will do a lot of smal edits so it isn't lost again >(

 

OK, I made a pic since this is something newbs often don't undertsan but is very important. I myself had quite a time figureing out what to do so hopefully this helps.

 

Turning Edges

 

Fig.1 is a sphere. as you can tell by the red lines, when a sphere is created all the triangles run in the same direction. \

It's possible in the southern hemisphere they run in the other direction /

 

This will render fine IF the sphere is unchanged as all the squares are planar (the red edges became visible when the sphere was triangulated, they split the original squares in half)

 

Also in Fig 1 you will notice that I shaped the sphere a bit. I grabbed the red verts, pulled them up a bit and rotated them foward. This gives it more of a lute body shape, or a pear shape. But it makes some of the square polys non-planar. This is where turning edges comes in to play.

 

Again, Fig 1, there is also a green and yellow line (yellow can't be seen in this view but is marked by an arrow).

 

Fig 2 and 3 show how these 2 lines make the shape look from the side. You want both sides to look the same but one of the edges is wrong. You need to turn an edge to make both sides of the shape the same.

 

In Fig 4 the yellow edge has been turned so it is symetrical with the green edge. Now that part of the shape looks the same from all angles. You'll also notice that the red/blue dotted line has been turned from it's original position in Fig 1. This is important to do throughout the model to make it symetrical. The edges in between don't matter as the polys are still planar.

 

Also in Fig 4 there is a pink line. it doesn't match the white edge. The white edge needs to be turned so it is aligned with the pink one. On this shape it isn't critical, but it's a good habit to get into making symetrical objects. It also illustrates that turning an edge is just aligning that edge with different verts to create a different shape.

[Baddcog]

@demagogue

thanks for posting that, I didn't realize LW counted only visible polys.

for the record, 3ds Max counts all triangles as polys (even if the object hasn't been triangulated) so an 800 poly object will laways be 800 polys.

It also doesn't count verts OR lines as polys. It must be a triangle at least. It counts verts as verts. A line would just be 2 verts.

 

@Springheel

I did try to convert bumps to normals with photoshop/Nvidia filter. Bad results though. All the sharp edges were beveled (at least one pixel) thus mis-aligning the bumps with the diffuse map. On some models it's probably fine, or at least being aware of these issues while painting the bump and diffuse will help.

[Springheel]

All the sharp edges were beveled (at least one pixel)

 

You're right, normalmaps can't do sharp edges per se. But like I said, I've heard D3 converts heightmaps to normalmaps anyway. If so, there's not much difference. I can't verify that it's true, but I have heard it more than once.

[Baddcog]

well, the model I am refering to was my lantern model. It deffinately looks different with a bump than it did with the normals.

 

I have also read that conversion thing but it might be unsubstantiated. I have several objects with bump and they deffinately look like they should in game, unlike when I tried to convert thru photoshop.

 

I'm actually curious if D3 turns twosided materials into double poly count. Also something hard to tell. BSP did this with Thief objects during conversion if the DBL flag was used on a material. For awhile I thought I was creating lower poly stuff than I was, bummed when I realized what was happening.

[Maximius]

But like I said, I've heard D3 converts heightmaps to normalmaps anyway. If so, there's not much difference. I can't verify that it's true, but I have heard it more than once.

 

 

I just read this somewhere in the D3 texturing Wiki a day ago or so.

[demagogue]

Today I was just doing a little brush up on the design texture for my lute, and when I saved it in Photoshop the file somehow got corrupted. I was keeping backups when I was making it, but there was this little window of time after I deleted past backups and before I made the next one. So now I lose like a week's worth of work before the last backup ... grr. (Essentially I used a photo source of a perfect looking arabesque hole, but I had to erase the strings covering the hole in the photo ... pixel by pixel). A little demoralizing, but I'll go back and do it over again, since I'm basically almost done with it, and it seems much, much faster now that I know what I'm doing with it. Just wanted to blow off some steam.

[Maximius]

I have a question about Blender. Im making a vase and I cannot seal off the ends of it. I don't know how to:

 

1. Select individual points, not just edges, if this is possible.

 

2. Select an area and fill it automatically, if possible.

 

Here is a picture to demonstrate what I'm talking about.

 

http://www.UploadYourImages.com/view/70145...nder_vase_1.jpg

 

 

I want to fill in or join those points or something that are currently selected, create polys in the middle, whatever. If this is not the best way to make the object, I'd appreciate some pointers, but I've always found this way best in LightWave.

 

Blender is starting to get a little fun to use, its really keyboard driven and with that keyboard map in the Wiki Im gettting more comfortable. It seems a lot more intuitive than LightWave. I would definitely recommend any new users to give Blender a try first and then tangle with LightWave later if you want.

[sparhawk]

You must enter vertex editing mode (select model and press TAB), then unselect all vertices. Select the vertices that you want to fill and press f to create a face. Hope that is what you meant. I don't know of a way to fill a big area with faces.

 

If you start a model from scratch, what I usually do (because it is less cumbersome then the above), I create a cube or some other object that fits the shape, but it doesn't really matter. Then I select a few vertices, where I want to expand, and press e for extrude. This will already fill in the faces. I usually create a few extra faces this way, to have enough vertices to work with. When I have to much I can still delete them, and then create the missing faces on these few manually.

 

As for Blender vs. Lightwave. I havent used Lightwave yet, but I really love Blender. It's quite fast to work with, because of the keyboard shortcuts, and it does the job very well.

[OrbWeaver]

I just read this somewhere in the D3 texturing Wiki a day ago or so.

 

Yes, this is correct. The Doom 3 rendering algorithm uses normal maps, so height maps have to be converted first (this is what the "depth" parameter in the heightmap() declaration is for).

[Springheel]

I really love Blender. It's quite fast to work with, because of the keyboard shortcuts,

 

Is there a modeling program that *doesn't* have keyboard shortcuts?

[Maximius]

Of course LightWave does but it seems to still be much more menu driven. And more, theres just a feel to Blender that is somehow smoother, more friendly once you get past its somewhat odd initial features. Its hard to say exactly why. For some reason things seem to make more sense.

 

When I started Lightwave, I struggled mightily with it. Nothing seemed intuitive, not one bit. I finally got going and still learning new things was a bit of a chore, even stuff that later seemed really easy and straightforward. Ive only been on Blender a month or so and yet I am already feeling a lot more comfortable every time I use it. Im hardly proficient but it can feel it coming a lot faster than with LW. Learning LW first of course helped in a general way, but not enough to account for it all.

 

Also, it doesn't hog system resources like Lightwave does, it seems anyway. I can run Blender, GIMP, Firefox w/o a hitch. For Lightwave I had to shut down all security and the web, boot it up, then slowly open things back up. If I missed a step, Lightwave would lock up in mid boot up and I would have to restart the puter to get it to work.

 

Im sure you can do a lot with LW you cannot do with Blender but no big deal, I'm not that good yet anyway. Blender is definitely the way to go for the noobie modeler IMHO.

[demagogue]

Well, so far I'm sold on LW.

So far, I don't get the crashes or slow downs.

I guess I've just gotten used to it after go through a few tutorials.

Maybe Blender would be better to start with, for learning, but since I'm so deep into LW now there's no good reason to turn back.

Also, I did a search of the models coming out for each one, and I noticed some of my favorite artists tended to use LW. So that was enough for me.

Really, I am so sure at this point that it doesn't matter, though. For the kinds of models we're making.

I'll just stick with what I've been using.

 

 

As for the interface, I think it matches the way I think.

The basic menu-structure is what you very basically want to do: create points/surfaces from scratch, move points/surfaces already there, create/delete new geometry based on what's already there (like beveling), etc. -- and then you go into more detail from there.

 

I like thinking about it from very fundamentally -- what do you fundamentally want to do -- to more and more detailed as you go on. And then most everything has its own sub-menu to do things numerically.

 

I think it's true, though, that you really have to explore all the different ins and outs of a series of menus before you use a function, e.g., just experimenting with each tab on the sub-menu. Then once you've seen it, you know the path to get back to it. It's hard to do something from scratch, before you've experimented with it on your own.

Edited September 10, 2007 by demagogue

[Maximius]

It may be too that while LW was literally one of the first software programs I've ever tried to seriously study( I usually learn what I need to and end it there with office platforms etc.) learning Blender feels easier because its not the first time. Both in terms of knowing what you are doing and thinking you can do it and do stuff with it. If you get my drift. To make a vase in LW, I had to learn LW and learn how to construct a vase, or anything else. Now, I have good ideas already in place for how to make those objects and can simply proceed with the technical problems.

 

If I can get this other HD installed, and get some real memory going, Im probably going to install both. But still, Blender feels lighter somehow. Its good to have access to both of course, learning one has synergy with learning another.

 

BTW here is some software I will be working with in my course. Its called NetLogo, we are using it to construct simple scientific models. I have not played around with it yet though myself, got too much reading right now.

 

Edit: Duh, forgot the link.

 

http://ccl.northwestern.edu/netlogo/

Edited September 11, 2007 by Maximius