Stage 2: Translating your concept into a 3D model
Advise for starters
There are two ways how you can model characters. You can either use the Box modelling or the Poly modelling method.
Box modelling means that you start with a simple cube, add a few cuts to block out the rough form of your character and have the proportions correct from the beginning on. Only after the form looks right you will go into the process of detailing the mesh.
Poly modelling is a different method where you extrude poly by poly or edge by edge to get the desired result. Most artists mix these two methods while modelling anyway, so do not limit yourself to just one method. Nevertheless most game character artist prefer Box modelling because you do not have to concentrate on proportions and details at the same time as you would have to with poly modelling. Others begin to block out the body parts with simple primitives, like a cylinder for the arm, and join them afterwards to one single mesh.
The Pose
Surely you have seen a few wireframes of game character artists in the internet and you surely have noticed that they all are modelled in a similar pose. That pose is called the T-Pose, because if you see someone standing in the distance in that pose you will see the 'T'. His arms resemble the horizontal line and his body represents the vertical line of the 'T'. This pose will make the rigging process easier because it is symmetrical.
Your character does not have to exactly stand straight with the arms in 90°. The important thing about that pose is that your character appears to be relaxed, which makes him or her look more naturally. Personally I prefer to slightly spread the legs and rotate the feet accordingly as well as lowering the arms a bit and moving the forearms slightly forward. I think this pose looks more humane because no body would pose like a T in real life without being seized with cramp.
The Form
Getting the basic shape right is the most important thing and I think the hardest part in the modelling process. Try to express your form with only a few vertices without any details, like for example the nose. You sort of try to draw the profile of your character and to build the silhouette. Remember that you should not just add new loops without tweaking the vertices. Beginners often tend to add new loops even if they could have gotten the same shape just by moving the already existing vertices to the right place, which would have saved many senseless polies.
The basic form of a head.
For practising, it is always a good idea to repeat this step a few times to get a feeling for the form. This step is also essential for creating your base mesh you can later use for sculpting in Zbrush, Mudbox, etc., but more about that later. You can find more information and nice images here:
http://subdivisionmodeling.com/forums/showthread.php?t=482
Poles
Some readers might ask at this point why we have to bother about the right topology and poles if we are passing our model into Zbrush nonetheless. I added this topic here because not everyone has a computer that is good enough to display millions of polies. Personally, I have not been able to surpass the 700,000 polycount on my machine. Although knowledge about poles is especially needed for subdivision modelling you can gain valuable information which can help you with your low poly work.
If you look closely at wireframe shots of various characters of professional artists you will notice two types of poles: Those with 5 edges connected to one vertex and those with only 3 edges. SomeArtist, a member of the subdivisionmodeling board who created the best threads about modelling theory I have read so far, called the vertex with 5 edges an E-Pole because whenever you extrude a quad you will end up with four E-Poles and four N-Poles. That is why the E stands for Extrude-Pole. The N is the short form for Nose-Pole, because this type of pole can be found nearly on every nose models. The N-Pole gives you greater control than the E-Pole simply because fewer edges end in one vertex. When box-modelling a head you basically hide one N-Pole inside the eye and another inside the ear.
The E-Pole and the N-Pole.
The 'Nose-Pole'.
Meshes without poles have to contain n-gons or are completely flat like a plane. The Doo-Sabin algorithm subdivides meshes so that an E-Pole is converted to an n-gon and an N-Pole becomes a triangle. This subdivision algorithm is quite uncommon because it destroys your uv layout. Poleless models have no real advantage over model with poles, so do not be afraid to use poles.
The wrong placement of poles can make it difficult to tweak a mesh and can lead to undesired bumpiness by applying the Catmull-Clark algorithm, which is essential for sculpting.
Undesired bumpiness.
Keep in mind that bumpiness is not always a bad thing, because we humans do not have a perfectly smooth skin. If you feel that a pole should be there, than do that. Poles on flat surfaces are fine anyway, so you do not have to bother about them in some areas. If you have stumbled upon a pole which is in the wrong place you can easily shift it around by creating or deleting edges, etc. There are endless possibilities for moving poles around. Just experiment a bit. Good examples about shifting poles can be found here:
http://subdivisionmodeling.com/forums/showthread.php?t=907
Flow
Flow is very important for meshes that deform which very likely your game character should do. The direction in which your edges seem to run to decides if your model is going to deform properly or not. Bad deforming models usually have a bad flow. Be especially careful when you are modelling a head and extrude some faces. Extrusion will always lead to a circular flow, which is in the case of the human head very bad. Circular loops are formed by two E-Poles on the same lane.
In general you can make a distinction between key loops and fill loops. Key loops are defined by poles whereas fill loops lie in between key loops. Fill loops are necessary for increasing the resolution of your model. That is why it is a good idea to go into detailing your mesh only when you have proper (key) loops.
The eye region is a special case because the eye is the most important feature of every character which determines if your character is going to sell or not. That is the region of the face which should get the most of your attention. Newbies simply extrude the eye area, which requires no knowledge of poles but will not lead to optimal results. Instead of an extrude loop most professionals go for a spiral loop because it looks more organic.
The baisc flow of a head.
The lower your poly count is, the more you will likely go with a cylindrical approach for the shoulders. If your engine allows you more polygons you choose the route to model out the muscles. Try to always think about how your mesh is going to be when modelling.
Sculpting
Now comes the fun part: Sculpting. Your base mesh should consist only of quads because most programs have problems initializing the symmetry when you have triangles or n-gons in your mesh. Split your models up so that you can use the full polycount limit of your hardware for every body part which will be far more in the end as if you would have imported one single mesh. As a positive side effect this keeps you more focused on every single area of your mesh. Import your parts into your sculpting application of choice and go insane with details. At this part anatomy knowledge is of course crucial.
Re-topologising
Re-topologising is the step where you create a new low poly model and by 'tracing' your high polygon mesh.Try to match the surface of the high poly mesh as closely as possible. The goal of this process is to capture the silhouette of your high poly mesh and express it with fewer polies. The low poly mesh you are going to create is the final mesh which will be visible in your game. Make sure that the poles are placed in the correct place and that your model bends properly. During this process it is very important that you also respect the polycount of your game engine. Nexuiz character models have approximately 3,500 polygons. For a list of polycounts of various games look here:
http://www.rsart.co.uk/2007/08/27/yes-but-how-many-polygons/
Re-topologising a mesh in Blender.
Here are a few tips for re-topologising meshes:
1) Odd sided cylinders work better for tubes and any other round object that you won't see the ends of.
2) Three sides for very small things, 5 for medium, 7 for large and 9-15 for cylinders that show the end caps.
3) Odd sided objects don't get that squash stretch effect when they rotate like even sided do. They will however wobble but this is something to only worry about when doing tires or other such things.
Where to go from here
If you still have problems modelling characters you can always ask in forums and teach yourself with books or DVDs. You will surely learn a lot even by watching other artists modelling characters, especially the time lapses of Bay Raitt, who did the blendshapes for Gollum and Martin Krol who is a very talented artist and the alike. Most videos can be found on youtube or simply by searching via google.
Edit1: Added a few images.
just change it using "playermodel" command in console and then "chase_active 1" to go into third person mode
