Eight questions from a newbie graphic designer on creating 3d models for computer games
Hello. I address the following questions with a view to a computer game for the PC (RPG from isometric projection). This should help you a little bit with answering my questions.
- Can one model penetrate with its topology through another model or should I combine them into one? What are the advantages of both solutions? Pictures to help interpret this question: https://imgur.com/a/xAsSOkH
- I have come across information that most of today's low poly model games use bevels with one segment to help baking. Is it true?
- Does doing bake in marmoset below 4k make sense? I got the impression that there is a big difference between 2k and 4k bakes and it doesn't look good qualitatively. I'd like to add that I just started learning marmoset baking so maybe I'm just doing something wrong. I have had to remodel my models so many times for incorrect baking, do you have any hints for that?
- If I do bake in marmoset and I intend to texturing in substance painter, is the only thing I need to import into substance painter is normal map?
- How can I best determine if I have overdone the number of polygons of my model? ? When I notice that the optimization of my game is not very good, should I start with optimizing my models or have a predetermined limit, e.g. for the location of 100,000 polygons?
- Should every model have its' own texture? If I remember correctly I read somewhere that one texture can be used multiple times, should I avoid painting manualy texturing then? My hand-painted elements will pass to other models in completely different places, Am I right?
- I'm having a very hard time figuring out which texturing method I should use. When do you use individual texturing, tiling textures or trim sheets? I would be grateful with the examples of use because as I mentioned it is hard for me to understand, I'm not even sure if I understand these concepts well. Perhaps I have to worry about it unnecessarily at this stage of my studies. By the way, if I only relied on one texturing method, would that be a big problem for optimization?
- Should I worry about where I put my seams when creating UV? A lot of people advised me to do it in invisible places, but to be honest to me, seams are invisible and I have never seen a clear cut in a texture. Am I the only one who can't see it? Is it worth using this method still? (I hope it is not a matter of my eyesight, because I do not wear glasses that are too strong so as not to strain my eyes)
Replies
1: You can penetrate geometry like that, The issues might be you waste texture space if the model is uniquely unwrapped and baked. zfighting could also occur depending on scale and how far from camera it might be (like pretty far though so not a huge concern)
2: Probably not true across the entire spectrum of games and types of assets in games. no bevels + hard edges are probably more common. Partly because the art in the majoity of games is not AAA realistic stuff. Partly because people prefer to minimise gradients in normal maps for a multple reasons. Adding a bevel and soft shading has it's place in baked assets, as does mid poly weighted normals, etc, Will probably find bevels and baked normals on AAA fps game weapons that are shoved right up to the camera fill a lot of the view and so on, but that's a very tiny niche of games as a whole.
3: There is a probably a noticable difference from 2k to 4k in a lot of instances. Context matters though.
4: bake all the maps substance needs not just normal for best results. Generating some maps from the LP and the baked normal is noticably inferior to baking it from a high poly.
5: Just follow a general idea to not waste polygons and testing/profiling your game.
6: depends on context.
7: You'll probably have to use tiling textures, trims and do some things with shaders in engine to tint or blend textures together. It'll just be much faster more versatile and allow you to make changes or entirely new assets at times as the needs of the level dictate.
8: put seams in logical places would be good practice. You'll find that hard edges and UV seams are not perfect and that tiling textures over beveled mid polys would also show up seams more or less depending on the texture.
7:. You work out what to do based on texel density which is directly connected to the size of the object and how you lay out UVs. If the object is too big to fit in a reasonable sized texture (2048*2048) then you need to think about tiling / UV tricks.
It's fairly safe to assume that 1024 pixels = 1 metre for an FPS game.
People will be along shortly to tell you that 4096*4096 textures are fine or that we can use 8k maps for everything because unreal5 but these people are wrong, can't do simple maths and don't have a real grasp on how hardware works.
With that said.
If you're new to this, do whatever allows you to complete your work . don't agonise over what's correct or most performant, just get stuff made. Once you know how to make stuff, you can worry about how to make it properly - there are too many things to think about and too many context specific things to consider and you will end up confused and frustrated.
Tldr. Do what feels sensible until someone tells you that you fucked up