What is the sense of use 3D and convert to 2D for textures?

santa_oscuro

First of all sorry for my english XD The thing is that I've been studying 3D from few years ago, but i never understood a thing and now that I'm traying hard with UV-mapping I thought it was time to find an answer. As you can saw at tittle, what is the sense of work with 3D and create 2D mapping texture?  Why the hell uv-mapping has to be done instead of making a texture map in 3D? It turns something simple into wasting time and probably in wasting resources.

I've been looking for the answer on the internet, but I didn't find anyone asking or answering for this logical question. Maybe is that I didn't know how to search it properly (although it seems strange to me), but  I hope someone can give me a satisfactory enough answer.

Thanks in advance.

Eric Chadwick

Quality.

For example
https://polycount.com/discussion/comment/1916391/#Comment_1916391

Alex_J

I thought method of uv unwrap came about because that was how technology developed? 3d painting didn't come until later. 

Still lots of good reasons for texturing in 2d especially regarding tiling textures.

Eric Chadwick

BIGTIMEMASTER said:

I thought method of uv unwrap came about because that was how technology developed? 3d painting didn't come until later. 

Not really. Inefficient UV mapping is bad for several reasons. This is coming up on 8 years old (8!) but still very valid.
https://polycount.com/discussion/85675/uv-unwrapping-best-practice-factors-priorities

santa_oscuro

Eric Chadwick said:

Quality.

For example
https://polycount.com/discussion/comment/1916391/#Comment_1916391

I don't undesrtand why add a new dimension on UV map or  another way of proposing 3D texturing can reduce the quality. If you don't do things correctly, of course that it will be easier for the quality be affected, but otherwise it shouldn't be reduced. Sure that optimization would be affected but facing the future it should not be a problem.

Noors

Somehow that's a funny question, but i guess uv are not that intuitive.

Well we need a support for storing the color data, right ? Something like a pixel grid !

What about a 3d grid of pixels ? (aka voxels)

3d textures exist. A 256pixels 3d texture is basically 256 2d textures of 256*256 pixels.

256*256*256 is a lot of pixels.

So there you are, your object stands inside this 3d pixel grid. But you only need the color information on the model surface, right, like a skin. All that pixel volume, just to fill your model surface. What a waste of data. And what if your object has to deform and the texture has to follow the deformation ?

So 3d textures are costly and used for special effects that requires volume (fog/cloud for instance). They are usually procedurally made aka based on mathematical equations (like a perlin noise) and not edited manually.

 UV "describes" only your mesh surface, which basically is flat triangles all around (yes a triangle is always flat ) UV are 2d coordinates and they are pretty light to the hardware. If the model is stretched, the texture will look stretched. Rendering pipeline are built upon vertex/faces for decades now, so it's the most efficient pipeline.

So no, it's not a waste. But unfolding might be boring/time consuming i give you that.

I'm not sure what you have in mind. Something like Disney Ptex ? But while you don't have to unfold manually, there are still uvs, and color is still stored in a texture map. It's just invisible to the artist. And while Ptex is valid for movies, it's terribly unoptimized for a real time engine.

Something like zbrush polypaint ? There each vertex is the support of a color information (basically it's called vertex color). So if you want fine details, you need billions of vertices. Fine for sculpting, not optimized for rendering, deadly for realtime. So you need to bake them to 2d maps aswell.

Shaders can create uv on the fly. Something like triplanar mapping could (should) be used on terrains, some assets...

But it's still heavier to calculate than a mesh with uv data. And you don't have much control over it either.

So basically, learn to unwrap...

But tools are getting better, to a point where unwrapping can almost be fully automated. AI will help artists, if it doesn't destroy humanity first.

RyanB

You can use constructive solid geometry or voxels if you want something with an "interior" volume.  Then you can use a procedural or volumetric shader.  I started out using POV-ray and did everything with constructive solid geometry and 3D shaders.

Alex_J

@Noors nailed what I was getting at. Dinnit write my question that great. Will read Eric's links as well.