Hey guys, I have created a big list of items that I want to model in order to give myself some practice before I go into 3d sculpting and poly painting, and I was wondering if I could have assistance with a few things.
Propane tank (low-poly):
Questions:
1. For an average game engine such as the Source, Unreal, or CryEngine, what are some good poly counts for things like props, characters, etc?
2. What is the rule of thumb for choosing texture sizes, and
how can you lower the resolution after creating a high res one?
3. Does 2000 triangles for a 3d propane tank (possibly view-model) seem reasonable?
4. In my picture of the tank I modeled, have I properly seperated elements from eachother to be exploded for baking? What are some tips you guys can give to me so that I can avoid baking errors?
5. How do you explode models for baking with keyframes? I cannot find any information on this, and I just remember reading it somewhere, and learning that it makes everything easier because you can just revert it back to normal when done.
Thanks
Replies
Instead of exploding try assigning material ids in 3ds max. Ex: high poly piece 1 wth mat id 1, lowpoly piece 1 with mat id 1.This can is better forme to bake rather than exploding.
I forgot to mention that I use Xnormal for my baking workflow as I find it better than the built in 3ds max baker.
Would this material id's thing still work?
Also, regarding question 1, I took the time to look through some Half Life 2 models with the hl model viewer in Source SDK. It sort of gave me the idea of the poly count for certain world models. Characters tend to go into the 10k+ range, while props stay in and around the <1000 poly range. Large building assets/view models go somewhere in between. I actually found a propane tank that was only 400 or so polys...
I also wanted to toss another quick question into the mix:
See the holes on my propane tank? Can those be achieved with an alpha channel texture? How would it work, and is it better to just get hard geo into the model as opposed to using a texture to fake a hole?
I don't know how studios work in general, but I think it's common to have to go through everything afterwards and do an optimization pass once you have everything in the game. I think it's pretty hard to visualize how many verts you've got on screen when you're working on one prop at a time.
What I think you might want to aim for is (1) make sure it looks good, artistically. And (2) make sure it makes sense, and balances out. For example, in your tank, you've got lots of verts in the fine details, but very few rounding the forms at the base of the tank. Would you really need those verts when you're looking at the tank in your screenshot view? If you lost some verts in the details, and rounded the bottom out with them, would anyone notice the loss in details? Probably not. However, if your game revolved around the top of the tank, then it'd make sense for there to be denser verts there than at the bottom!
Just go through your loops and see if any of them aren't really contributing in terms of form, UVs or baking, and that the vert distribution makes sense for whatever game camera through which you intense to view it.
But if you're practicing as if you were on the job doing the real thing, take into consideration all the other variables that may be at work, such that having an alpha channel in your texture actually doubles your textures' memory size, giving less room to other textures in the game that might be more important than a propane tank >D
Considering that, you may want to just stick with the actual geometry. Nowadays, hardware can take the hit of a ton of polygons with practically no sweat, unless you're working with mobile devices of-course.
Organised studios will do a memory plan for levels and assets. So there will be vertex count limits worked out in advance for various kinds of props.
This is why it's more helpful to make a complete scene rather than a single prop. You can get a feel for how to organize workflow for memory and performance in advance.