3D painting in Blender an alternative to Alpha Mod?

[Ferrao10]

Maybe this is the wrong place to ask this question but I could imagine some other people might get interested as well. I already registered at Katsbits, waiting for approval now. In the meantime I thought I'd post here. If this idea has already been thought through years ago, I'm sorry. Didn't follow the scene a long time.


A colleague of mine has been working himself into Max and Mudbox over the last couiple years and sometimes we talk about his progress.
He once told me about this fantastic, intuitive way of 3D painting in Mudbox and now I wonder if this would be an alternative using Blender compared to Alpha Mod texturing in Radiant.
AM seems like quite a hassle to me from Socks tutorial(s). I have to admit that I myself didn't even try to apply it to something yet.

My aim would be a canyon/cave map with quite a lot of different textures fading here and there.
I read most of the tutorials over at Katsbits and got me Blender and started playing around with it, like, 14 days ago. So, yeah, you can call me an absolute noob and maybe naive as well .
From what I got from the tuts and my own experiences; using meshes in Blender and then exporting to .map would result in Thin Skin brushes which are not acceptabel to work with in such project.
OK, nurb surfaces would be exported as a mesh. Cool, but I don't get the idea behind nurbs yet. Don't bother, my colleague could probably explain that to me and show me the basics right inside Blender.
But wouldn't it be better/more comfortable anyway to export the "terrain" in chunks as .md3 models?
Having the seems be where VIS would make the cuts, using caulk/hint-walls and clipping like you would with a regular terrain?

Yeah, those are my basic thoughts.
Questions are:
-wouldn't 3D painting be more comfortable/intuitive than AM texturing?
-would I be able to export whatever "already textured material" from Blender into something that I could build upon in Radiant?

[skinNCNmaster]

if youre saying that you make a "skin" stye texture from a top down view of your up facing polygons, on say a 1024/ or 512/ size jpg, andthen scale it soit covers te entire area you're attending to, then yes, your idea is sound and very strong 'trick" when it comes to getting a animated feel.

like this thingy:


something like the modelskin style thing; the areas all mapped out so you can paint one large texture to cover different ground brushes.... this ones a rather non-detailed one
[lvlshot]http://www.digitaldrone.com/-sanchez/RuinEntrance.jpg[/lvlshot]

you wouldnt make the texture like thisone though due to this one being used on multiple faces of say a non conforming brush shape, ie, a staircase side has a zig zag alpha, it ons any face to look like its been cut into staircase shape, when in fact a staircase can be faked with a recessed ramp thats textured to have stairs..

the same thing is relatively easy to do in gtk, make a solid top down texture of anything you want, highly detailed art to match the other stuff you plan on putting in, after you've got to use a projector to stick it to your func_groups, ie different top down textures for different spots.. sometime you might apply the texture manually, but if your brushwork is angular on anything but perfect 45's you'll have aton of rotating and nudging to get things aligns nice.

or a simple modification to something that exists..., say this is the top down view of an outdoor area... all it needs are finer details for the flow routes, add in some footpaths and stuff... paint your shadows right onto it.. the fine detailed ones.. do you see how it sort of defines the geo it wants for its hull?


this image though is a broad area image, not your fine detailed style one which would only cover a small area..

[Ferrao10]

I don't have a plan about UVW. Look at me as if I were a child. I could come up with some idea noone has had before because I'm not thinking in restricted ways. And mabe that is bad. Maybe it doesn't work out at all as I had thought it would.
I have to learn a lot about modelling. Lke I said, I just been using Blender for 14 days now...

My final plan would not be achievable with a height map and meta shader as I understand you suggested. Because I'm thinking about "positive" and "negative" volumes within one map. A normal terrain (as in Team Arena) does not have a "ceiling" and so it is possible to draw it with a heightmap. But once you have a ceiling (a cave, a rock-bridge, etc.) it's not possible to work with a heightmap+metashader anymore (or say, it would be more practical to use some other form of editing when you don't want to split the hull into different segments).

I also don't want to hassle around with func_groups and projectors, either.

My sole questions are already outlined pretty clearly in my starter, imo.

Here's a graphic for illustration:



Yellow would be gravel/dirt, green grass, blue rock, purple some other rock, etc.


And edit: Skin, have you ever worked with Blender and exported something practical to Radiant?

[skinNCNmaster]

well its true you could do the work in blender, i dont know much about that specific program
i do get modelling pretty good though, I've used alias wavefront power animator v8(mayas parent) and wings3d(the simplest editor ever created, and one of the best free ones) and gtkradiant,..



what your modelling in blender is not very complicated to do in gtkradiant. I could replicate that geometry in gtk in less than a half hour i guess..

now your idea is to do something like this: at least this is how i would achieve it were i thinking along the lines of your 'canyon" being a repeating tile on the ground and sides, but not a tile which repeats on the sides/top of the bridge, the trick is you get the textures in a single large file and make sure you can put the edges up against one another, this means extending the image data from one texture onto the other..

you need to keep in mind is this technique is about making the entire side of the stone bridge in one texture image, and then fitting the texture to the bridge width... you would aint the entire scene in a rectangle and leave put the same part in both images use for the "overlap" areas..
sorta akin to skybox making, the "corner seams" connecting three different image files at a shoulder/armpit or elbow



1. func_groups in gtkradiant just means you grab all the brushes (solid geometry shapes) and call them one piece, like your color coding, projecting just means you put a camera entity and face it down at those brushes and tell it to paint your texture on them from projection instead of individual facial coordinates.. thus youget one big image and tell it to hit all the brushes you want it on via projection... that is one way, it is mainly for near flat areas, very low range angles, when things get off a certain angular point in gtkradiant the textures twist 45 degrees.. so we tell the texture how to be dealt with by the engine in a script. we specify the angular direction of the "projection" either using a camera connection of one entity to a target position or by just sticking the texture to the brushes in any setting at all(it can be upsidedown and fit 1X1000) the SCRIPT or shader tells the q3map2, the rendering compiler how to place them, and the game engine how to draw/align them.

although, yours looks relatively simple to achieve with a couple of large wide textures, the work that you put into the imagery is the key, highly detailed things.... and having optional versions of the similar theme.. such as having two or three versions of the same texture with major modifications to everything but the edges...

if you are interested i can help you via a msnm client to learn some of the basics to get you up and running in gtkradiant... i mean sculpting things like your model there... i've got a pretty solid grasp of how to with gtk... (pardon me if i mistake you for who knows naught/little about gtk.. cause there are many tips/tricks that are not in tutorials...

now i know this is neither of your choices... not alphamod or model, its a sort of trick, start with a base texture tiling on all textures that will have touch edges, it may be that in some places with this method you only need a detail to cover the meeting place of a couple different texture corners...
0

[Ferrao10]

I could be wrong here, but doesn't 3D painting mean you paint the whole surface with whatever textures you want like with layers in Photoshop?
So you don't need to bother about tiling. At least not to an extent you couldn't compensate for manually?

I know that geometry is not very difficult to do in Radiant. (Actually it is a worked on export). But Radiant tends to corrupt brushes on vertex editing, blender does not; at all.

[skinNCNmaster]

3d painting means you export your geometry to a square file, it becomes a series of lines like the shape of a cardboard box before folding it up, that is Not what im saying to do, but it is one method. or its the ability to use a paintbrush which "draws" 3d geometry as you move your mouse, like in maya, you can choose "vines" and then paint actually 3d wire frames just by dragging the cursor across a preexisting face or facet of geometry. OR it means you have paint type tools in a 3d program and you spray paint color onto the model, then export that work to the same square file...

im saying, think, your ground is a canyon/valley, which is nont enclosed on one side or the other, so there WILL be tiling of your "wide" textures, if you consider your canyon walls continue on... your road continues on.. you do not want to get up into huge image sizes...

so im saying you can texture that entire scene with 5 images which you paint in photoshop using a combonation of compositing and painting.

[obsidian]

Skinmaster, Ferrao10's not talking about UVW-wrapping or other texture projection techniques, he's talking about texture blending. If you're going to reply, please don't interject with superfluous stuff not pertinent to this topic. Read his questions and if you don't understand them, posting irrelevant stuff will just make it more confusing for him.



Definition of Terms

"3D painting" is confusing and can refer to a number of different things. If your colleague is using Mudbox, he might be doing 3d-sculpting, which involves "painting" in details to a heightmap or triangle tessellation. This technique is usually used to produce models with millions of polygons which may be "baked" down to a normalmap. However, this has nothing to do with texture blending on terrain.

I assume you mean, "vertex paint". This allows you to assign vertex colours and alpha in your modelling program to the vertexes of the mesh by "painting" in values. Yes, it's fast, easy, intuitive and you get to see what you paint, but it does have a few difficulties.

Difficulties With Vertex Alpha Models

You can paint vertex alpha values directly onto your mesh in the modelling program and Q3Map2 can process these vertex alpha values, the big caveat lies in getting your modelling program to export these values in a format that supports them.

Q3Map2 prefers the .ase format, but the .ase format only supports vertex colours, but not vertex alpha. I'm sure .obj supports vertex alpha, but I have never been able to get materials to export properly with .obj. I have no experience with .lwo and .3ds. Anyway, point is that it will take a lot of experimenting just to get the model out of your modelling program and into GtkRadiant/Q3Map2 properly. It's probably doable, but I haven't been able to figure out the process myself using 3ds Max. If you happen to have better luck, please share your results.

Alternative #1: Q3Map2_fs

Q3Map2_fs is an alternate branch of the main Q3Map2 project. It is experimental and has no official status. But what is interesting is that by design, it "incorrectly" processes vertex colour values as vertex alpha. So you could paint on vertex colour information, export to .ase and compile with Q3Map2_fs and it will convert the vertex colours into vertex alpha. While the results work well, I don't like using Q3Map2_fs from a purist's standpoint, some of the "fixes" in Q3Map2_fs are a bit like disassembling your car's engine to power your toaster because you blew a fuse in your electrical box this morning. But metaphorically speaking, at least you'll have your toast.

Alternative #2: Just Use AlphaMod Brushes

It's not that hard, trust me. The process seems complicated, but it's actually quite easy in practice. The hardest part is setting up your blend shaders for the terrain, but you have to do that regardless of whether you use alphaMod brushes or vertex alpha painting. Sock uses little "pegs" for alphaMod volumes, but on much of your example mesh, you can use a single brush. For example, you can envelope all the vertexes of that yellow mesh with a single alphaMod volume brush.

[skinNCNmaster]

Skinmaster, Ferrao10's not talking about UVW-wrapping or other texture projection techniques, he's talking about texture blending. If you're going to reply, please don't interject with superfluous stuff not pertinent to this topic. Read his questions and if you don't understand them, posting irrelevant stuff will just make it more confusing for him.

stop attacking me and reply to the thread poster, you dont know if he wil be confused or not FOR him do you? You are out of line friend, I'm contributing a SUBJECT ORIENTED RESPONSE, your job here is to help and remove people who are out of line, i am providing game editor/subject oriented information and if it is not from the angle you want it to be from you descriminate against it?

obsidian you are beginning to become a very bad administrator around here. BIG BROTHER my ability to provide someone else with a half hour of my time... my life and the time i spend trying to help are worth less than your replies? STOP coming in with a "THATS not right, this is... mentality'

here are two side tiling textures i made but have used yet for similar type of usage... a smaller chunk of canyon wall..

[lvlshot]http://www.rave.ca/en/image/full/378679/[/lvlshot]

what you would do is make sure these have corners that are edited from the ajoining textures


and although obsidians options are some of those that exist, he is not god in this arena there are more than 2 ways to skin a map. we all know different things, and though i have not personally completed a map using the suggestion i provided, it is not a impossible nor is it a bad suggestion.

[obsidian]

I could be wrong here, but doesn't 3D painting mean you paint the whole surface with whatever textures you want like with layers in Photoshop?

If you happen to be talking about painting the texture directly onto the model, consider the relatively large amount of texture data required to cover the surface of a terrain map, every pixel would have to be unique across the entire surface of the terrain. This will not work well for id Tech 3 given the limited memory constraints of the engine and the GPU's VRAM. Newer engines like id Tech 5 use "megatextures" to compress and stream that level of data to the GPU, older engines simply don't have that kind of technology.

Otherwise, if you're just talking about assigning textures to the surface by painting them on the blend areas, see post above.

[Ferrao10]

Sorry to have brought up misconceptions.
I thought "sculpting" and "painting" were some totally common and, by now, discrete terms. I mean isn't it something totally different, already by the name?

I'm going to read on to your comments.
Sorry for any trouble...

[skinNCNmaster]

remember what he is telling you about is "blending" that is having a nice continueous edge of transition from any one texture to another, what im talking about is taking a long look at the brushes/geometry you've got as the "colored" area, and in a rectangular image file in photoshop, "painting the entire thing using image compositing/every trick in the ps book.. (ie, lots of variety, and optional secondary textures, which are similar with detail changes, for anything that is likely to tile.... then say you wat a path, you only need copy the canyon wall texture, crop it to the a narrow width and paint n a path, then change the canyon wall brushes were your path is to wear the modifeid version (with the path painted on it)

that way you can "fit" that texture to the entire colored region... (this works bet if you don't have to scale the texture up too may times..0

obsidian official sayer of right and wrong solutions. ..!!!?./!? not the solution "I'd" choose, or not the "right/left soluton lol

its a wonder there arse oaf you people in this forum these days

[obsidian]

Whoa, you mad bro?

If I wanted to troll you, I would have edited your post with photos of your meth stained teeth or just delete your threads entirely. It would certainly be easier than criticizing your posts, that requires an actual effort on my part. Believe me, I have better things to do if I did not think that there was some benefit to my comments.

I'm not saying your "method" won't work, I'm saying it's not the right solution to the problem he is describing. For the technical reason why this isn't what he wants, see my post above. For a metaphorical reason, refer to the toaster.

[Ferrao10]

Wow, had to take a pee.


I thought breaking the whole .md3 up into severeal chunks (individual .md3s) woud help as well with texture scaling problems.
Like that graphic I brought up was about 1/5th of the middle of a 'CTF map on Y-axxis. Would that be helpful?

Obs, I do thank a lot for the info. Maybe I should go through that AM- process even once.

[obsidian]

Clearing up some other confusions:

There is a difference between a heightmap and an alphamap. You seem to use the terms interchangeably.

Heightmap is a black and white image that describes to a program differences in elevation on a segmented mesh. Quake 3 doesn't use heightmaps in any way, if you're reading about old-school Team Arena terrain and the Terrain Manual that ships with some versions of GtkRadiant, the heightmap is only used by the GenSurf plugin to generate the terrain geometry. Most people will choose to manually trisoup brushes or model terrain meshes.

An alphamap is an image file that "maps" the vertex alpha values of a terrain mesh. id Software used an alphamap for the terrain of Team Arena. It's really finicky and complicated to use and as you realized, it only works when you have a trisouped terrain with no overhangs because it is projected along the z-axis. This method is pretty much obsolete since ydnar introduced a way to manually assign vertex alpha values using alphaMod volume blending.


What is Vertex Alpha?

A vertex is a point in 3d space described by it's coordinates. In computer graphics, a vertex point can also contain other data, including colour and alpha values. So if you look at a terrain mesh, in addition to defining the shape, the vertex also describes colour and alpha values at each vertex point on the mesh.

Quake 3 can read the vertex alpha values for each vertex on the mesh to modify the transparency of a stage of the shader. If the vertex has a v-alpha value of 0.5, the texture will be assigned a 50% transparency value at that point.


How Do I Assign Vertex Alpha Values?

By default, vertex alpha values for every surface of the entire map is assigned a value of 1 (no transparency). To set values, there are a few ways of doing this:

1. Alphamap - used by Team Arena. Hard to use, has limitations, deprecated.
2. q3map_alphaMod dotproduct - auto-assigns vertex alpha values depending on vertex normals. Typically, the greater the slope of the normal the greater the transparency. Works great for stuff like snow on a mountain.
3. q3map_alphaMod volume - manually assign vertex alpha values by enclosing the mesh's vertexes with a "control brush", which sets fixed values for every vertex inside of the "control brush". See Sock's tutorial. This is the preferred method, or used in conjunction with alphaMod dotproduct.
4. Vertex paint - manually assign vertex alpha values by "painting" values on the mesh's vertexes. Supposed to work, but exported model formats are troublesome.

[phantazm11]

The .lwo format can handle Vertex Alpha blending through Vertex Painting. I haven't had trouble with the .lwo format and q3map2 (it is all I use) and I believe Blender can export the format, though I haven't actually used Blender created .lwo files.

[Kat]

Check your mail Ferrao, you should have got an auth request. If not let me know and I'll activate you manually.

It been a long time since I did anything with Quake 3 but you're best bet would be to look into vertex blending, certainly for the type of object you posted an image of, although that looks like a simple object, when you break it down into shapes that need to be textured it's actually quite a complex form if you want textures to orientate themselves correctly.

Blender 2.49 can definitely export painted ASE, not sure about LWO (which are a bit tricky to work with in Blender). There is a new ASE exporter for the latest versions of Blender but I've not done any vertex painting with it yet. And yes you'd need to split the models up anyway if it go big enough (you might need to do this anyway based on material assignments as ASE can only handle one at a time - LWO can do more but as mentioned above, they are tricky to export from Blender).

[Theftbot]

Multi material assignment on ase is already possible using 2.49 and goofos exporter.

[Kat]

Well yes, but that's not what is meant by my comment above; there's a difference between exporting a single object with several materials associated with it, versus an object composed of several materials and sub-objects. One works, the other doesn't.

[Ferrao10]

deleted on User.
Kat, check your PM.

[dONKEY]

Hey Kat:)
Slightly off topic, I got the ase exporter working in 1.51, is there an importer that works?

[Kat]

As far as I know, not for the new versions, you'll basically have to import using the 2.49 and then reopen the file into the new to get the material into the app.

@Ferrao: back at ya, check PM

[sock]

AM seems like quite a hassle to me from Socks tutorial(s). I have to admit that I myself didn't even try to apply it to something yet. My aim would be a canyon/cave map with quite a lot of different textures fading here and there.

Creating terrain for Q3 maps is a two stage process, creating the actual terrain (brushwork in GTK editor or modelling package and ASE/LWO export) and controlling the blend shaders on the terrain (alpha fade brushwork or vertex painted with a LWO file).

Creating terrain can be easier in a modelling package but several things need to be considered:

* Make sure you get the scale right (width/height) for the player. One of the great advantages of the GTK editor is when you apply textures you get a sense of scale. Convert a chunk of brushwork in the editor to a ASE file and import into your modelling package as a reference. (Donkey does this well in his tutorial - http://www.leveldk.co.uk/q3tut_model.htm)
* Scaling models can be done in the editor but I would not advice it for floor stuff. (This problem can get worse if you use model with collision/clipping enabled)
* Q3 Models do not have collision by default and it has to be enabled on the model entity (in the editor, spawnflag) or manually done with brushwork. The preferred method is a simple brushwork hull because complex model collision will cause problems with bsp and bot files later on.
* Never make floor models complex because players always hate being thrown around the place because of uneven crazy floor designs. Either make sure the floor is clipped smooth manually (in the editor) or make sure the central paths through the terrain are smooth in the model.

The terrain blending can be as simple or complex as you want, depending on how satisfied you are with the end results. I know people moan about how I use alpha poles in my terrain to create blends but this is not necessary. The blends can easily be done with huge brushes in the editor as well. The reason I do this method is so that I can see in the editor where the blends are happening (there is no visualization of blends in the editor) and then placement of additional detail models is easier (plants/rocks etc)

Kat, check your PM.

Got to say this is a shame, discussing this in private messages. I am sure there are many posters/lurkers here that would benefit from an open discussion of how you are getting on with your modelling terrain process.

[obsidian]

I think their PM has to do with registration issues on Kat's forum, not about actual level design or modelling stuff.

[Kat]

Yarrr it was. And double yarrrr to generally encouraging public discussion about Blender and terrain generally.. yarrr!

[Ferrao10]

Sorry for answering a bit late.
Hadn't had much time lately because of the family.
Read all through the post again and realized I made confusing comments. I'm sorry for that. I'm new to everything after the simple heightmap+alphamap+metashader technique. I hardly know the correct terms for most of what came after those days.
First, a little about myself, so you may get a clearer impression:
I have experience in Radiant. Started to map back in 2002, I think it was. Then had a break from 2005/6 till 2010.
I'm a man with little time and several strong interests, and a family with kids.
That, Sock, is why this project of mine probably (or rather surely) won't progress anytime soon. Sorry, folks, don't expect any screenshots in the next future. I only have an hour or two a week for this hobby. And I have to learn blender and AlphaMod.
That also is the reason why I ask those generell questions before I simply test it myself. I don't want to naively run into something that isn't even doable wasting that little "free time" I have for myself.
Yep, I had some problems registering at Katsbits, that was the PM. Kat, because of the above reasons I haven't started a post over at yours, yet.

I didn't know the term was "vertex painting" because my colleague always called it "3D painting". From what I understand, with this technique you can use a virtual paintbrush or airbrush to paint on a model inside blender but with a texture applied to the brush instead of a simple colour. So, I'd take a base texture (in my case the tileable rock texture that would cover most of the surface) and apply that to everything. Then for example, I'd start to paint in the grass using a tiling texture, then the dirt with a tiling texture again, and so on. The brushes I choose act like brushes in photoshop. They automatically blend the edges and on top of that I can control the opacity of the paint/texture as well as brush size and so on. In my understanding I don't have to write a blend shader because the model would be exported with this one texture I already painted onto it; that already contains the blends I need. Kat, you mentioned texture orientation on models. Do I have to orientate them manually? Or are they mapped by the adjacent normals?
/// Next part contains pieces that I'm really unsure of, namely modeling:
Up until this point, till I have painted the whole object/surface, it is one body.
Now I have to cut it in chunks, already keeping in mind where I'd place my caulk-walls and hints. Because no origin should end up inside a caulk-wall; because of and in regards to VIS; and because of texture-sizes loaded into the GPUs memory. Am I restricted to 512*512 textures on models/md3s?
Next up I'd have to convert each chunk into a nonsolid md3.
/// End of work in blender
In Radiant I'd now have to import the md3s and make sure they all sit perfectly alligned to each other. No sparklies, just perfect fit.
Then I'd add the caulk-walls, hints, clips, areaportals, lights and so on.

Would I run into seems between the individual md3 chunks because of the export-import?