I'm planning my first UDK environment and I've been all over the place looking at pieces and different workflows. Right now I'm specially interested on modular modelling for buildings facades.
On the subject of workflows one thing that's not clear for me, about the more realistic type of objects (walls, windows, doors, bricks) what's the trade of of using just photos to get normals against sculpting them myself ?
I think sculpting in general takes more time, and I'm not very good at it either. But this can be an opportunity to get better, If there's a really significant reason to do it.
Replies
In any case, if it's faster to get and use photo source normals then do it, sculpt your own if you have the time afterwards, it may turn out you only need so many normals than what is originally thought. So why do more assumed work when you won't use it all?
A lot of games use photo sourced images and use programs like crazy bump or xnormal in photoshop or substance to generate their normals, they may then go in and adjust them afterwars if needed, but you're not a production team, so just go with what keeps you moving to the end goal.
You can polish once you're near the end. You should.
if sculpting takes more times and you are not good a it, then definitely focus that, to get faster and get better at it .
I certainly would not say this. Physically accurate shading has little to nothing to do with how you create your normal map content. If the content is good content, its good content, if its poorly done, its poorly done.
I use a mix of modeled, sculpted, hand painted height, and photo-dirivied normals.
Everything has its place depending on the type of material you're creating. I wouldn't try to create normal maps for complex geometric/mechanical shapes from photos, and I also wouldn't waste my type painstakingly sculpting fine patterns for fabrics, concrete(aside from large/medium form shapes/wear) etc.
Your PBR shader has no idea where your normal map came from and will not scold you if you didn't create it in some specific manner.
The sad truth is, that real-time rendering is nowhere near close to be able to pick all the small details.
You shouldn't use big fakes, but small ones are ok.
As for normal. It doesn't matter. As long as you get the result you want. I would still try to learn sculpting some things, as you will sooner than later meet situation where you can't just find right photo, for what you want. And only options left are painting heightmap in photoshop (have fun) or using zbrush.
Not disagreeing with you. From what i've seen how the albedo textures look it's probably just gonna be a case of running a couple more filters and adjustments to remove most of the lights and shadows.
And you are probably right that some of it will remain. When the last gen of materials came around people said the same, that diffuse maps only should contain color information yet ao maps and fake ao stuck because it simply looks better.