Texture Creation: Packing more into each pixel
polycounter lvl 17
Alright. I have been texuring for some time. And I guess I am in need to a refresher or good base tutorial.
I am working on a dumpster, and while doing so I ran into this:
Where did this guy learn to cram so much detail into each pixel. My texture was looking alright, but the more I stare at his he seems to be making -every- pixel go to work for him. micro scratches and tiny leaks. There just must be something I am not getting, perhaps even on a fundamental level. I guess it's time to relearn how to do this crap.
I'll post mine so I can learn and be ripped a new one alike
I haven't started in on the grey areas yet.
/discouraged
I am working on a dumpster, and while doing so I ran into this:
Where did this guy learn to cram so much detail into each pixel. My texture was looking alright, but the more I stare at his he seems to be making -every- pixel go to work for him. micro scratches and tiny leaks. There just must be something I am not getting, perhaps even on a fundamental level. I guess it's time to relearn how to do this crap.
I'll post mine so I can learn and be ripped a new one alike
I haven't started in on the grey areas yet./discouraged
Replies
The title of your thread is the solution to your problem. And you answered your own question: Don't be intimidated to go in and spend time with a tiny brush and add scratches/detail.
yeah that's a 2k texture for a dumpster.. pretty ridiculous
2. When you bring in textures for overlays do you sharpen them before applying them?
3. "depth or detail"; I get detail, what do you mean by depth?
I may just have blown a gasket today and become so dismayed that I don't feel I know what I am doing anymore.
I think my original 512 was a little too low, and thus looked bad. I may give this another shot soon.
I dunno why I hadn't thought of that.
http://cg.tutsplus.com/tutorials/autodesk-3ds-max/how-to-create-a-video-game-dumpster-the-complete-current-gen-workflow/
I am going to watch the texturing portion of all that I hope I -don't- learn anything. I am feeling pretty incompetent now and doubting my own abilities, but I may just need rebake my normals an AO at 1024 and start over.
What dogma leads to people choosing the colors they do. Some choose odd colors, many choose blue or orange, and some don't color their specs at all.
-I- have been deriving my colors from my Diffuse, blurred, and high contrast. I just need to know what others are considering when they do theirs.
And you should look at the whole tutorial, your mesh looks a bit funky as it is now and I would suggest re-making it all over again, that is the fastest learning method anyway.
Also this guy have some pretty awesome tutorials on how to paint realistic metal, it should also be able to help you a bit.
http://www.racer445.com/
As a matter of fact, I did make some changes to the mesh, and to the high-poly as well. Over all I like the shape. If I can find the time, I may just watch them all.
Those other tutorials are great too.
Usually the ones who don't color their spec maps are programmers, lazy, or can't due to engine constraints (source engine in particular comes to mind.)
Specular colors are crucial for defining materials. Lets say for instance you have 3 kinds of metal on your item. On your spec you could make one of them yellowed, one blued, and the other a green. This will help separate them from each other. DO NOT just make everything lazily blue!
example:
Sometimes spec colors need to be extreme to sell materials. These rounds are a good example.
Something I was taught recently was that dark diffuse + bright spec makes things look shinier than what most people tend to do (bright diffuse and bright spec.) Someone even showed me an example where the diffuse was almost pitch black and the specular was doing all the work, and it looked great!
On that note, specular also needs to have its own detail that only shows up in the light. This is important to make your materials more dynamic. This this example, the shots on the left have no light on them and thus look fairly flat, while the shots on the right have light on them and you can see a good amount of unique detail appearing. The above shot of the texture map also shows some of the actual detail.
The key to metal is to KEEP IT SMOOTH! So many people just lazily throw 1000 overlays on it and call done without thinking how metal acts in the real world. Those people cant use speed as an excuse either, because adding all those overlays take more time than just picking a few really good looking details and using those. Less really is more in this case.
I hope this helps you understand proper specular usage for metals more.
ps: gloss is amazing too, but lots of times we dont have the budget for it. usually people just do flat color values, but in practice it can be so much more...
EDIT: It may have been environment lighting or something like that.
I don't know if I am being nit picky here or just reading wrong, but I don't cosider specularity related to reflectivity in any way other than just how reflective the object may be. True, specular is the feedback of light off the model given a direction, but you can't really get the real specular of an object through reflectivity, and you can't do the vice versa either. I hope I make sense.
one thing that comes to mind is like, gray trash cans of plastic that tend to be stretch cast or blown, resulting in artifacts that are visible usually only in light. These in a cloudy day or in shadows can look dull or solid with no specular (barely), while on a sunny day, and facing the right directions (more to specular than directional, but bear with me) you will see the grooves and scrapes of the plastic shining a different pattern than you see beyond the specular (or in a cloudy day with the can in the shadows)
Specular is used with the purpose of showing how hard or soft/matte the surface of an object is, true, very matte things don't reflect, but the reflection is still separate phenomenon, or only related, not the same, and definitely not a way to fake reflections. Hope this helps.
Boths shots, 1 photometric area light, shape visible to renderer.
Phong Shader, no raytraced reflections, specular on.
Phong Shader, Reflections enabled, specular off.
Notice both has the same specular hotspots
NinthJake: it may have been cubemaps he used. that would make sense
Spec helps us fake awesome reflections. Games don't have awesome reflections right now, they have cubemap reflections & spec. If you had awesome reflections, you could control intrinsic properties of how light is absorbed, reflected&refracted, surface roughness & light diffusion, and many other things.
But, in general, we don't do that. We make generalities about how it is very reflective and mirror-like and it gets cubemaps, or its rough and not mirror-like and we use spec & gloss to fake it. we can talk for hours about the subtleties and techniques we can use to make these two cheap methods very convincing, but the bottom line is that SPEC IS FAKE GUYS. accept it.
And furthermore learn to understand it because a lot of people use spec & reflections in inordinate proportion, and the same is true with glossiness & reflections. Learning how this stuff really works, in real life, will help you make your materials more convincing.
I don't either and you're absolutely right. Most of the time it seems to just be some silly shader requirement.
In some cases (generally wood with coatings or rubber/plastics) you DO need to separate them as those materials pick up a soft diffused white "specular" highlight and thus look bad with much of a cubemap reflection. However, these instances are fairly rare and in many cases you can still get away with using the same map for both.
I really can't wait until realtime raytracing is in place. :P
[ame]
Granted, if the asset is moving around like a car or character, it can be expensive to resample the cube every frame (or even sample multiple statics, like HL2), but if you pull the perf from somewhere else, it does make a noticeable difference in looks.
One thing that is very important to note, is that unless you're packing grayscale spec and some other maps(gloss, etc) into other channels, you'll receive absolutely no performance benifit over color spec.
8 Bit grayscale spec maps can not be compressed, so they end up being the same size in memory as a full color texture. So unless you're doing something weird to optimize, or using uncompressed textures, there generally is absolutely no reason *not* to use colored spec.
Not to mention that subtle or even extreme color differences is a great way to sell different material types.
Specularity means mirror like, you have a specular or diffuse reflection depending on the surface.(in realtime shaders diffusion is handled by gloss)
Being a specular reflection in reality has nothing to do with not reflecting the environment. It just means that the reflection is not being heavily diffused Otherwise im totally with you. using cubemaps to simulate ambient light/bleed makes things much more believeable.
This is kind of going off on a tangent though
The issue i am having is how i want to retain ALL of the detail that i put into the model and keep it looking nice without it looking muddy. I want to start with a 8k texture but that is so stupid and overkill...and this model I want to show it off inside of UE3 (using UDK). What is a good method for this? Could I just have the texture at 4k and scale it in the engine? (also using unity 3d) I don't want to mirror parts of the armour since this is a boss character and want him to have asymmetric damage.
Thanks, you guys are the best.