Do you prefer accurate ware detail or accurate reflectivity?
polycounter lvl 11
I've been looking through some of my favorite models on Artstation and I've noticed a pattern. I tend to enjoy models that have more accurate reflectivity value and lighting, even if the surface detail and ware/grunge/grease does not look as accurate or scales incorrectly. Am I alone in thinking this or is it a common thing?
Replies
Reflectivity, or more precisely Roughness, has a lot to do with perceived realism in a surface material.
for example, i saw in a documentary, sometimes naughty dog will tweak reflectance values to make materials pop or chill out a bit depending on the lighting situation. for example, boosting or lowering the reflection of puddles in dark corners to make them not as glaringly bright.
sometimes exaggerating say the reflectiveness of a metal that is next to a more matte surface can really help create a better sense of material definition, but generally these are small subtle tweaks based on art direction.
Else roughness / reflectivity!
Also it depends on the over all artistic goal of the game.
But yes, different people will have different artistic goals in mind, some want to go super hardcore on the realism, others have no problem screwing with the lightning and material properties to get a better result, most settle somewhere in between.
EarthQuake wrote:
This question makes no sense, accurate wear and accurate reflectance values are not mutually exclusive.
+1
While I'm at it, reflectivity does not = roughness. Reflectivity is the intensity of the reflection, be it diffuse or specular reflection intensity. Roughness or gloss represents the microstructure of the material, and determines how smooth or rough the material it is. The smoother it, is, the glossier or tighter the specular highlight. The rougher it is, the more the light is reflected in random directions, which results in a wider highlight. Roughness does not modify the reflectivity or intensity of the highlight, but simply the behavior of it. Rougher highlights will appear less intense, but this is not because they are less reflective, it's because the light is spread out over a larger area (energy conservation).
A: Have accurate roughness detail (or just believable shading in general) and half-ass some default Substance Designer noise nodes
B: Really make sure the ware detail (scuffing, scratches, dirt, etc) look believable and whip up some half-assed adjustment nodes on the overall roughness/metal value?
Again, you can only choose to make ONE look accurate.
If anything I would recommend you to simply practice texturing rather than trying to imagine such scenarios.
Also, wear.
What makes PBR so cool is that the system is 99% designed with real world physics in mind. Every material should actually behave the way it's suppose to, it's just the artist's job to help guide the computer in the right path.
However, I think if I removed all of that, and spent more time making the roughness values look really good, then this piece would at least be passable despite the unrealistic "ware" detail.
Even Naughty Dog's material artist said that she spends a lot of time making the "cleanest" possible version shade as accurate as possible before adding any ware detail.
Seriously! this 'topic' is an oxymoronic contradiction, if ever there was one.
daniellooartist wrote:
Let's use the phrase "ware wear shape" and "ware wear size." Easier?
...and no.
P.S.
A word to the wise, as at @pior alluded too, why not simply test your peculiar summations whereby in due course results often, well at least in my experience tend to speak for themselves?
daniellooartist wrote:
So here's an older piece I did. You might be thinking to yourself "holy shit that thing looks like it's been through a nuclear holocaust." That's because I put on all the dirt, all the scratches, all the grundge, all the things!! That's what I meant by ware.
However, I think if I removed all of that, and spent more time making the roughness values look really good, then this piece would at least be passable despite the unrealistic "ware" detail.
Geez...an adjective correctly spelt as WEAR
...plus this walkthrough might help you out Game Prop Texturing Fundamentals
Anyway I'm done.
Adios
But even still, going overboard with detail is considered noise, which is not really accurate but an amateur mistake.
It's like people who are new to normal mapping that slap on any texture because it's "bumpy". Even though veterans know this is wrong and they actually put the time into sculpting normal maps by hand and getting results that are more consistent.
I personally spend extra time just making sure all my material reflects correctly before I even start piling on noise/wear/whatever detail.
Throwing wear and effects on top of nonsensical base materials will not make a good texture. As you mention, this is something that is very commonly seen in student work. People work really hard to make sure the texture seems interesting in the 2D map, without regard for how it actually responds to light. That said, it's something a lot of pros are guilty of too.
Your question is very general as I believe there is alot of difference between accurate visualization for product presentation/ development and game assets. I dont think they are the same thing. But to answer your question I prefer substance so far as aspects of the materials are more flexible. That is personal.
This way of asking ("do you prefer this over that ?") is doomed from the start because it is an attempt to force people into saying what what you want to hear, even involuntarily. Also asking "which one do you do first ?" is just as problematic, because it assumes that people do one after another, which is just another assumption pulled out of thin air.
Now imagine the kind of answers you would have gotten, had you simply asked something along the lines of :
- - - - -
"Hello, I am having trouble with texturing this asset (pair of headphones). Despite my best attempts, the W-E-A-R and tear seems way too strong, and the material properties seem way off. Here are all my attempts at fixing the problem (pic 1) (pic 2) (pic 3)". I am using X and Y softW-A-R-E. What are your suggestions ?"
- - - -
As long as you're following the correct PBR pipeline, it takes more effort to intentionally break the system then let it run its course.
Both the detail and the lighting are important. Yes, I understand that. But both take time to make and test. The shape of your "mask" that represents the detail and the "uniform color" that represents your surface lighting (using Substance Design terminology) do in fact exist independently of each other.
Ideally you want to make both look good. Ideally. Which one is more important and which one should be dedicated slightly more time in the event of a non-ideal time-crunch appears to be a matter of opinion based on your responses. I honest to god can't tell if that's what you actually believe or if I'm just failing to articulate my question.
Like I mentioned before, my eyes are very sensitive to surface lighting and my mind will fixate the slightest inaccuracy. I believe accurate surface lighting bothers me more than it bothers you guys. I've already uploaded an image 3 posts up to make my point and I honestly don't know how I can make this any more clear. If this is, in fact, a non-concern as you mentioned, then I will just keep doing what I normally do.
This is probably the assumption that derails the whole discussion. If both seem to take time to "make and test", then there is a fundamental issue with the workflow being used. Metalness is pretty much on/off, Roughness is a 1D scale from 0 to 100, and wear and tear can be set temporarily using a generic mask. This baseline can take as little as a few seconds to establish per material.
If it takes more than that, then that simply means that you are probably tackling the issue in a overly complicated way - which doesn't have much to do with "which one to do first".
It depends on the art style, it depends on the specific material you're creating, it depends on the engine and shaders and light and and and. Really, I've never in my career had to decide between good looking base materials vs nice detail and wear, even during crunch. I might spend a bit less effort on all aspects if I need to pump something out quickly, but you can't skip any of the fundamental steps.
As Pior says, if you're struggling to find the time for base materials and wear, your workflow is probably pretty messed up. Using Substance Designer may be part of the issue, as far as texturing unique assets go, something like Substance Painter, Quixel dDo or straight up Photoshop may be a more straightforward process. If you're inventing Substance Designer nodes from scratch to come up with your base materials and wear masks, you're probably wasting a lot of time.
Something that I find very helpful is creating simple parametric materials in Toolbag (you can use other tools or engines, but it's best if the lighting and shaders are representational of your target, which is something that is generally easy to accomplish with TB's modular shader editor). I can quickly enter in some values to block out the abledo, spec, gloss (or metal/rough), and flip through different HDRIs to see how the materials react to different lighting environments. It's trivial to make adjustments to these sort of parametric materials, so you can set up your base values in a matter of minutes. I like to do this process with my high poly asset before I get down to baking. It's good to see how the forms read with the base materials as I'm modeling, so I know if I want to fatten up a bevel or tweak a certain area or what have you. You can then bake these out as maps or just copy the values into whatever texture creation software you're using.
You can do the same sort of thing in Substance Painter and other apps if you've got a material ID mask, just make a simple fill layer for each material type and adjust the reflectance values until they match your reference.
Now, back to crunch. Thinking about it as X vs B is not productive. In reality, you should have a workflow that scales to the importance and/or time frame of the asset. Let's break it down into a few explicit steps:
- Base materials
- Wear
- Unique details
- Lighting and presentation
For the base material pass, at minimum you should be able to set up flat color values in a matter of minutes if you're pressed for time. If you have more time or the asset is very important, you can spend time researching the specific material (how it's made, it's reflectance properties, it's structure), and then finding or creating unique texture content (usually tiling maps) to represent that material really well.For the wear pass, you can scale it from running some preset masks, to building your own custom mask stack with grunge passes and hand painted elements.
For unique details, you might just skip this step if you don't have time or it's an unimportant asset. For hero assets, you'll spend time setting up unique details that tell a story and add character.
For lighting and presentation, it really depends on why you're making the asset. If you're building something for a project, you want to make sure your lighting matches what you'll see in your render target, so your asset looks as it should when it goes in. If you're creating something for your portfolio, you'll want to spend extra time setting up nice lighting, perhaps building a small scene to put the asset in context, etc.
So, how do you decide which areas of the workflow to invest more time in? Again, it depends. With more experience you'll learn to understand which material types are more difficult and need more time, and of course it depends on the art style, your specific goals, etc. The only person who can answer this question is you, and the only way to understand it better is to make a shitload of art and learn from your experiences.
Here's is the reference I used:
Here is the time I recorded for each step:
And here is the the prop I made:
Again, this was all rushed, but I never felt like making the actual roughness/reflection map was all that time consuming. I would actually say sculpting took the longest and more time should be prioritized on that. I basically ended up using 1 or 2 brushes in Mudbox just to save time and felt that hurt the results more. But everything else I made and expected the results to come out how they should (i.e, concrete, paint, rust and metal material).