Hi everyone. This is gonna be a really dumb question but here it is.. is the subd workflow still out there? what I mean by this is if the workflow of doing a low poly, adding supporting edges so that when it smooths it creates nice bevels BUT never actually doing a committed/permanent smoothing but instead a 'smoothing at render time' type of thing. I ask this because in the game industry you obviously cant do this, you have a determined number of polys and you go down from there with the LOD's but I don't know if subd's are used in the film industry. AND if it's still used what happens with the textures? because if you texture a low poly to be smoothed at render time some stretching of the UVs has to happen right? I hope I made myself clear. Thanks guys.
Replies
Games using Tesselation, which may or may not include the sub-d method you're referring to include-
Call of Duty: Ghosts
Deus Ex: Human Revolution
Metro: Last Light
Tomb Raider
Crysis 3
Max Payne 3
Battlefield 3
Alien Isolation/ AVP
STALKER
Dirt 2
~a whole bunch more~
Those mostly use a displacement driven tesselation, only a few use the Sub-d style smoothing, such as - the weapons and first person hands in Ghosts, Aliens in AVP, and character models in STALKER.
So the answer is yes, I mostly disagree with it being used without a map driving tesselation, but for creating a new ultra-close lod out of the lod0 asset that doesn't create any pop-in, there really is no other option.
I believe Call of Duty Ghost featured some dynamic subd, but I do not know if it was used beyond guns and characters. As far as I know, it hasn't been used outside of that game. Most games rely so heavily on quality baked normal maps, smoothing changes the mesh normals would break the normal map workflow. I assume it's also not commonly used for those stretching issues you brought up, and for performance reasons.
Tessellation is the closest you are going to get, which has it's own limitations, requirements, and uses, and isn't widely used, outside of terrain.
The problem of smooth UVS has been alleviated somewhat with open subdiv but, with dense base topology this isn't as much of an issue as it can be controlled with tightly spaced in-line edges, or not a problem at all in the case of the dense polymodeled meshes.
So far, the biggest pain has been the baking process for displacements and normal maps. Raycast baking has been mostly unsuccessful (organic shapes work better with raycast baking, hard surface has been a miss), but using a subdivision baking method works well. When the maps are baked using subdivision or the UVmatch feature in xNormal, then everything works.