Source engine character modeling question

k@rt

Hi All. I am working on character model for L4D2. My original plan was to use a single 2048 square texture space, so in the end I would have two 2048 textures for - diffuse and normal.

However my sculpt in Mudbox is such that if I do that I will lose some nice detail that I like, so I am wondering if its possible/worth it/feasible to up my texture resoultion.

My first question is if I were to unwrap my model to two 2048 textures and use a multi-subobject how much extra rendering cost will this generate once the model is in-game? It would have to load up four 2048 textures (2 diffuse and 2 bump) which is beginning to sound like a lot... but I really have no idea how "expensive" this is in real terms.

A comprimise could maybe to use one 2048sq texture and one 2048x1024 texture. If I unwrap my model so that SetID 1 covers all my 0-1 uv square, and SetID 2 covers just the top half of uv sqaure 1-2, and I extract the mudbox textures to that. Will it work if I then cut the second texture (uv square 1-2) in half (giving my 1024x2048 and then apply a uv xform to the SetID 2 polys to stretch them back out to a square? I am worried about seams not lining up very well.

In the end, is all this really worth it? I am wondering how much extra detail you would end up seeing on the screen.

Any thoughts would be appriciated.

cryrid

In the end, is all this really worth it? I am wondering how much extra detail you would end up seeing on the screen.

Might as well pop it in and see. It shouldn't take too much effort to save out a larger map and check it out in the engine. You'd get an idea of what it looks like in the engine, if you can really notice the extra resolution during gameplay, and what the impact on performance might be. If it's just a single model for a personal mod it probably wont make too big of a difference either way.

k@rt

cryrid wrote: »

Might as well pop it in and see. It shouldn't take too much effort to save out a larger map and check it out in the engine. You'd get an idea of what it looks like in the engine, if you can really notice the extra resolution during gameplay, and what the impact on performance might be. If it's just a single model for a personal mod it probably wont make too big of a difference either way.

The model isn't really "for" anything specific... I wanted the experience of modeling, sculpting and rigging a character model from start to finish, so i decided to make a custom "skin" for L4D2. But I want it to be as good as I can make it...

I take your point that I could obviously test all this. However re-arranging the uvs correctly still takes a bit of time... making sure they use the space as economically as possible. I'd need to come up with a simple diffuse, the do the test compiling, running in game, comparissons etc...

All of this isnt beyond me, it's just that this model has already taken much longer than I thought it would... and so far I have had to do a LOT of testing, experimenting etc...

I thought maybe somebody with more experience in modeling Source might have tried all this before and could let me benefit from their experience...

Thanks for the advice.

Angry Beaver

Characters in source tend to be made of at least 3 material sets anyways. One face, One eyes, One body. Source uses defined shaders so you're merely swapping out the materials themselves not the shaders so performance is generally decent, although it's not the best lot's expected for a character to eat up a lot and not outside the realm of being sensible.