This is one way to do it, it is not THE way to do it.
Also this is one way for now, things change fast. I reckon we will see the death of manual uv organizing pretty soon which will change things a lot. Having said that though,
1 - CREATION OF A BASE MESH (aka THE ROUGH SHAPE)
I have two options to create a base mesh: using a polygon modelling approach, through suites like Maya, 3D studio max or Modo or go directly to a sculpting program like Zbrush or Mudbox.
The first approach is perfect to obtain good, deformable mesh which are needed for animation and games but with the advent of retopology plugins we can jump directly to the second approach, which gives to the artist a more similar approach to classical sculpture.
At the very beginning you can decide if the scale of your work is important. Like you said if you are only going to make an illustration which will stay in the world of the one application you are working on then scale wouldn’t be important. If you were making a model for film or games then size will play a big part.
We use sculpting software because it is faster than any other method. We dont use sculpting for all 3d work but I only have info on that process. You can start your work off in any number of ways and you can ensure that the scale of the model is optimized for your sculpting app as well as any other software you will use. I am only experienced in zbrush (zb) but the principles are probably the same with other software.
You can start off in any number of ways: zb by using zspheres, or simply pull everything out of a sphere with dynamesh or make a base mesh in another app, but in the very beginning once you have a full size version of what you are going to make roughly laid out then normalize it (zb unify) so that the object is at optimal size and position for the sculpting app. Export that rough layout to whatever other 3d app you will use as well and ensure it comes in at the exact size and orientation you want for your end result.
The rough base mesh or stub, does not have to have optimal edge flow at all.
2 - SCULPTING
In this phase I sculpt the mesh until I reach the desired level of detail.
When I reach that level of detail, I use a retopology tool (like 3dcoat) in order to obtain good edge flow and polygon distribution (aka a 3d object which can be easily passed to the rigging and animation department; and, overall, an object with a better, more coherent “shape”).
Well not really. At this stage you can go crazy and do whatever you want until your computer stops. The hi rez sculpt just has to look good it doesn’t have to do anything else. You only retopo if the mesh is so high rez that you cant proceed further. This happens when you use extraction and insertion and all sorts of other tools that deliver very good visual results at very high polygon cost. Then you optimize your model and sub models, but more often than not you can just keep going until you are finished with the high rez. At this stage you paint your hi poly model.
3 - TEXTURING aka “PAINTING THE 3D OBJECT”
Here I start to have a lot of doubts (even because I’ve not reached this phase yet). With trillions of tutorials talking about projection painting, texture painting in photoshop, ptex etc I really don’t know what’s the mainstream purpose of this phase.
Let’s see if I got it correctly: to export my 3d object into Maya or Modo etc for rendering, I have to export the object and the texture file. I can’t export it as a unique “piece”.
In order to export the “colors” I have to put them in a texture file. I can create this texture files in various ways, the two mainstream ways are: painting directly the 3d object or painting the unwrapped model (a sort of model after a steamroller passed on it).
Yeah this is where it gets interesting and fun
You can decimate your high poly and export it at the right size and orientation (we did the preparation for this in step 1) to whatever app you prepared it for. The decimated version is made out of triangles and preserves most of the detail at a very reduced poly count if you are going to make your new mesh outside the sculpt program. If not just make a new mesh in zb (not the easiest, but in 3dcoat people seem to like the retopo tools). In any case you should see your hi rez as a template.
Whatever technique you use when making a new lower poly mesh by ‘drawing over’ your template there are things to consider. The new low poly must have good edge flow for deformation, it must be constructed in such a way as to optimize the use and production of maps for all the surface detail and it must have a poly count that reflects its end use, whatever that is. Once a good base mesh is made it must be checked for corrupt elements which you have to fix. Things like ‘t’ verts, overlapping and disconnected sub objects, well anything you can pick up with a mesh checker.
a) where do concepts like UV mapping fit in this process I’ve just described?
Once you have a good clean functional low(er) poly mesh you can organize the UVs. We can do all the technical stuff on a lower poly mesh , like uvs, skinning rigging and animating much quicker than on the hi rez version. So we uv this lower poly mesh. Just to confuse things though, there is yet another method artist use to export their models from sculpt packages into a 3d application with superior render capabilities. You can use auto uvs in the sculp app, extract maps straight away and export models and sub models to be rendered elsewhere. There is a good tutorial going from zbrush to max for rendering but I cant find it. At least you know that option is also available.
b) Why can’t I simply paint the 3d object and forget about everything?
Well you can just paint and forget. We now have two meshes: a hi with all the detail colour and shape wise and a blank lo mesh with all the functionality and a good set of uvs. You can extract maps using many different applications but I use xnormal because I can save my hi mesh on any disk and use xnormal to extract any maps I want without hasving to have any meshes open in any software. I point xnormal to the high poly and the low poly and tell it what maps I want depending on whether the model is for games or film. I think all other techniques involve having the models open in the application used for baking which means they mostly have to have the hi model reduced in poly count which gives me lower detail.
c) I’ve seen that surface details like pores, skin deformation etc are exported into normal or desplacement maps. This is just to take advantage of a low poly model “covered” by such maps right? I wouldn’t need such maps for, example, a 3d illustration?
d)What am I missing in this phase?
Yeah you are not missing anything this is correct.
The last phases go on with the rendering and compositing.
As you can see my problems are in the texturing phase; unfortunately all the info I found is old, broken in chunks or extremely jargon-rich. I’ve read a wonderful book “Digital Modelling” but still I have a lot of black holes. There is also the fact that there are infinite ways shown to reach the same result thus confusing things even more.
Well what you are attempting is to make an entire product which is what I like most. The disadvantage to this approach is it costs loads of time and effort across all the processes involved. These have become specialization areas worked on by teams and you want to do all of it. For example when I attacked rigging I woke up 3 months later and it was as if I had only been busy for a few days! That is what it is like with every part of the process. An excellent dvd collection going from concept to finished character can be found here: http://eat3d.com/blog/eat-3d/new-dvd-release-zbrush-35-character-production
This is a whole series going from start to finish with a few approaches to boot. It is very good and I am not affiliated with the product. Hope this helps. I know it is a lot and it takes quite a long time and a lot of effort to do everything but you will have a very good overview of the whole shooting match at the end. At a certain point the light will go on :lightbulb and never go out.