Why no layered materials?

[itsallgoode9]

I see what you're saying but I think we just have different workflows/render needs.

I see how that could slow things down. but so do SSS, Dispersion and Displacement...i'd assume much more so than a layered material. As it stands already, mix material is effectively doing the same thing...aka multiple materials being combined by layer masks. I already use the mix material in this fashion and i've had no issues, other than the system being cumbersome, limiting and confusing at times.

I'm sure it would take a good amount of code overhaul to implement this but that's a programmers job, to be honest.

I'm not a big fan of not implementing a feature because it's could slow things down. That should be left up to the user to decide, IMO. If i want to wait 6 hrs instead of 2 hrs because I added crazy SSS and Displacement, it should be my choice.

[grimm]

Yep anything you add to the kernel will slow things down, so choosing what you add needs to be thought about more to see how it will affect the overall user experience. Although with CUDA 7 I see that you can now dynamically compile your kernels. If Otoy can add this, it could help with adding more of these features. With this Octane could just compile a kernel that has those features you need and exclude the others. Of course this depends on how Octane is written and it might not work well at all, YMMV.

[fatrobotsneedlove]

I mean, that might be able to be done but you'd have to create extremely complicated roughness, spec and bump maps to achieve this. And once they are created, it would not be very friendly to make updates once you see they don't work perfect. ]

I guess you've never tried Mari? Or substance painter? This is exactly what they do, and they do it very effectively. You are able to create material 'layers' each with their own diffuse, roughness, spec, normal/displacement, etc that you have practically unlimited control over. Then you use the same tools to create layer masks. When you're finished, it flattens all of the layers using the masks into one extremely complicated diffuse, one roughness, one spec, one displacement, etc. But the project saves all of your layers, so it is actually very very easy to go back and make updates. Just have to re-export.

Not that what you're asking for is at all unreasonable. Especially in print, material layering is insanely common. I used to do procedural materials a lot when I used vray. It's just that it makes renders sooooo sloooowwww, so I found significantly more efficient means to accomplish the same thing.

[grimm]

3D-Coat does the same thing, make the needed changes, re-export, and Octane picks up the changes (or you can reload the parts that have changed).

[FrankPooleFloating]

dDo and nDo rock for this stuff too. Mari, Substance Designer, nDo/dDo.. take your pick bro (I personally picked dDo/nDo)... Fewer images & fewer items (one diffuse, one normal, one spec, one roughness etc..) to update whilst creating/tweaking. But for the record: I did vote yes. Always happy to have more options.. provided it doesn't slow shit down yet even more.

[itsallgoode9]

is there any way to see how these adjustments look in Octane while you're making adjustments in one of these programs? Or does it use a renderer inside of of those programs to get a rough idea, then you export a single texture out of of the program and import into Octane? I'm familiar with all of these programs but have never actually used them. If i have to export and reimport texture for every tiny slider adjustment, that would slowdown my workflow to a crawl. I'd never get a project finished doing that.

Also, I still don't get why the slowdown issue keeps getting brought up. If this is implemented and you you don't need a layered material, then don't use that node? Same thing as the mix material node now.

[grimm]

is there any way to see how these adjustments look in Octane while you're making adjustments in one of these programs? Or does it use a renderer inside of of those programs to get a rough idea, then you export a single texture out of of the program and import into Octane? I'm familiar with all of these programs but have never actually used them. If i have to export and reimport texture for every tiny slider adjustment, that would slowdown my workflow to a crawl. I'd never get a project finished doing that.

Also, I still don't get why the slowdown issue keeps getting brought up. If this is implemented and you you don't need a layered material, then don't use that node? Same thing as the mix material node now.

You can see the adjustments in real time in the plugins like the one for Blender. There are no plugins yet for these programs, that I know of anyway. My workflow is to, like you said, get it looking good in the program and then do the final minor tweaking in Octane. Any major issues would be taken care of in the external program, like UV's, decal placement, etc.

The slowness relates to GPU programming in particular. GPU's are designed for many concurrent processes that are very similar or the same. This allows the GPU to group like processes and memory accesses so they are very fast. The more features you add to the kernel code brings more complexity to what the GPU is processing. GPU's don't like that, they want all of the processes to finish at the same time, access memory at the same time, etc. So in general even if you don't use a particular feature it still affects the kernel as the choice of not using a feature disrupts the processing flow. The processors on a GPU are not like the more general purpose CPU's on your motherboard. They are much smaller and more integrated with each other and the GPU memory. A simple branch in a CPU is trivial, on the GPU it can cause all sorts of problems.

What makes Octane special is it's speed, which is amazing for an unbiased renderer. The first unbiased renderer I used was Indigo. I even built my own three node Beowulf system to run it on (6 cores). I could get the samples per second up to about 270K which is so slow today. Now I very rarely get less than 3 million s/sec on my 980 in Octane. But GPU's are just much more difficult to program than CPUs. Just look at VrayRT compared to Vray, not even close to the same number of features but what features you have are so much faster. Sorry for the long diatribe, hope it helps.

[itsallgoode9]

Thanks Grimm, that does help1 I knew that GPU renderers were different but I never knew exactly how.

[NVN]

the best and user-friendly material editor has Fryrender!
Why Otoy dont use the same concept???

to get the nearly the same speed, why cant octane translate this editor to his current materials before rendering
its only a user interface that will be translate???


for explain:

i have one material with 5 layered materials---->start Rendering-->octane translate this material to 5 Mix materials

Will Work???

in the attachment you can see my suggestion:

[itsallgoode9]

Since there was no response from the devs and nothing specifically was announced in V3, does this mean that there is no hope for a layered material node to be added?

Here's a real world example from a label I'm working on right now (my frustration is what lead me to revisit this post ). Graph node on the left is my current setup octane. Graph node on the right is a mockup what the exact same material would look like if a layered node existed. The mix material method takes 4 mix material nodes (which all nodes look identical in the graph making it extremely confusing). I really hope you can see how much easier the method on the right would be to work with (you can probably decipher that material without knowing anything more about it). Beyond being less visually confusing to work with, the layered materials are just simply a more efficient way of working. it's a nightmare creating materials when you can only combine two materials at a time. can you image trying to work in photoshop with this method??

any word from devs about where this stand; if it's even being considered?