A 110 thousand polygon head we will be re-topologising.
Re-topologising a mesh, that's one heck of a mouthful that sounds complex and scary to people new-ish to modelling. Really it's nothing of the kind, it's a simple concept that's easy to grasp and put into practice. Topology are the edge loops on your model, most people are aware that there are certain loops for example a head model should have such as loops around the eye's, mouth, nose / chin etc. Re-topologising is simply modelling with a total disregard for edge loops or 'good topology' and doing what ever you need to get the shape you require then remake the mesh with GOOD topology.
It sounds like a lot of work that can't make a lot of difference doesn't it? Lets imagine you've spent a while making a model in a displacement painting application such as Zbrush, Silo 2 or Mudbox. Even if you have good topology, you end up with a rather dense mesh when you're finished and if you wish to use or paint it in another application you're going to have to mess around with displacement maps. It's rare that a displacement map keeps the same quality of detail as the original sculpt, it can get damn close but not exactly the same.
The model I'll be using for this is something I'm still working on and is far from finished. I used it for two reasons, firstly because it was handy, and secondly that unlike human heads it was a little more challenging.
For this example I'll be using Maya, mainly as it's a no 'topology brush' workflow and the general principles are the same for most of the big applications. I'll outline the way I work as clearly as I can given the restrictions of both space and time.
We're going to set up a little fantasy project for ourselves where we have to create a highly detailed model that our pipeline needs to be able to later paint in Bodypaint (allowing us to paint all the maps at once) and this is then going to be sent to the riggers and animators. Obviously as the stand alone version of Bodypaint doesn't allow you to paint over a displaced model we may have some problems. Plus for all we know the renders for our 'fantasy' animation may be using a render engine that doesn't support displacement maps very well (or at all). But as our high polygon model will be done in an application such as Zbrush, Mudbox or Silo 2 we are going to have to re-topologise it to make it more 'pipeline friendly'.
What we're going to end up with is the same amount of detail but with much less polygons. This means shorter render times; less messing around setting up the scene and model for the riggers later, basically it saves valuable time in the long run.
What application you use to re-topologise your model is unimportant; it's the workflow itself that is the key. I'll show you a few ways of achieving it in a few different apps and this should then transfer well to whatever you're using in your pipeline.
Workflow - The Zbrush Part
Let's assume that our Dragon Head base model from Zbrush (or any other displacement painting app for that matter), has some bad topology, this can be a conscious way you've worked in another application while making the base mesh, or due to using Zsphere's in Zbrush and needing to re-topologise it for rigging.
We're go to try and to keep the detail in our sculpt using a simple and effective workflow. How far you refine this or take this method depends on the project and model itself. You don't want to do re-topologising on a very high dense mesh with fine detail (as this will be added through use of maps most times anyway), the aim is to reproduce the medium level detail well.
Firstly I took my base mesh that I knocked up and detailed it to the level I was happy with. Knowing that I'd be using a normal map or bump map with the final model I didn't take it all the way to fine wrinkles and skin pores stage. This means that once we've got our main forms and medium level detail on our sculpt we can the export our high level mesh. I also cut the model in half I Zbrush by hiding half the mesh and deleting the hidden side. This means as our dragon head is symmetrical that we are only going to have to do half the work (image 1).