To be more specific:
The (maybe nonworkable) idea is to be able to create a very high-res model in a sculpting application; and a very (as in very,very) coarse base model; and then let the tessellation and displacement do the rest.
Of an extremely high res model, from which you generate a normal/displacement map, which you can then apply to a coarse model. The other alternatives would be some form of subdivision scheme (e.g. pnTriangles, or subdivs). Maya/XSI allow you to model subdiv characters, and there are libs available to handle the tessellation (e.g. opensubdiv).
That should theoretically drastically shorten the time needed for the actual 3d modeling and rigging.
Or is this something the currently available RT tessellation methods are unable to accomplish?
It's entirely possible, but don't expect it to shorten the time for rigging & modelling, because it won't.
Ok, I guess that splits to two questions:
1)If there are, say, two characters on the screen - which will result in better visuals:
- using lowres meshes with tesselation and displacement map
- or using a hand-made high-res meshes, sans tesselation?
(With the same frame rate, of course)
From the stuff I've read and seen, I am still not sure whether the tesselation is only good for dynamic LOD (and as such, the most 'important' objects are still better left un-tesselated).
2)Why do you think it might not shorten the time for rigging and modeling?
Without tesselation, you need
- one very high res model to generate normal maps from
- one high res model to texture,rig and ultimately display
I though that with the use of tesselation, the second model can be considerably more low-res, and as such, easier to create and rig. Am I missing something?