[mjcherry]

Both Substance Designer and Substance Painter. I'm so excited about this! I'm to tired to do a write up now, but if anyone is interested, comment and I'll let you know what I've figured out so far.

[Mitsurugi]

I mean, there is a bunch of us interested, I am sure:) Let us know when you get some rest:)

[Mihai]

Any tutorial is welcome. You might want to contact NL through the customer portal and see if they would be interested in making a blog post about it. I'm sure they would be.

[lebulb]

I'm not sure what you're talking about,
i use maxwell and Substance Designer for years.
There is just a need for a gamma adjustment of the roughness map (if you use PBR)
and (or) to create speciphic channels for Maxwell.
The metal map is a rough interpretation of the ior (for PBR)

[mjcherry]

I'm not sure what you're talking about,
i use maxwell and Substance Designer for years.
There is just a need for a gamma adjustment of the roughness map (if you use PBR)
and (or) to create speciphic channels for Maxwell.
The metal map is a rough interpretation of the ior (for PBR)

Then we're both confused, because I've been researching this topic on and off for a month and have yet to see anyone describe a workflow that has worked for me (if they describe one at all other than to say where you might plug in some maps, which I thought was pretty obvious). Maybe I'm the only one who's late to the game though, I never claimed to be the brightest light on the street...

Your post is an example of one that confuses me. I am unable to use a gamma correct node with most maps using a Maxwell Material and Hypershade and the image adjustment parameters never seem to do the right job. Also, as far as I'm aware, you can't map the IOR value in maxwell to an image map?

My goal has been to get as close to a direct translation as possible between what we see in the iRay renderer and Maxwell. I understand the difference between the two render engines and am aware that there will be some difference (mostly for the better by using Maxwell) but still, if I cannot adequately visualize the work I'm doing in Substance and know that it's going to translate, the program is relatively worthless to me. The biggest hurdle (for me) has always been applying a dialectic material over a metal and getting them both to look right.

So here's what I did:

1) I constructed the Material for the Mole Fresnel in Substance Designer by creating two separate graphs, one for the underlying metal alloy layer (which I wanted to be old and dirty) and one for the paint layer. I then created a third graph to combine them.







2) While I'm sure there's a way to export the maps from Substance Designer, I don't know what it is, so I published the substance which I then imported into Substance Painter. From there I exported a full set of PBR maps for each material (alloy & pain) and the metallic map from the compbined material. I exported everything as EXR files as, after testing, I believe you get the most accurate translation that way:





3) I began building up the material in maxwell using these maps as you might imagine:

Basecolor -> Reflectance 0 & Reflectance 90
Roughness -> Roughness
Normal -> Normal

For the alloy layer this worked perfectly as expected however with the default Nd value it was way too bright. I then punched in the Nd value for iron but that was still too shiny as I wanted a rougher, older metal. I didn't want to have to do too much with the image controls and found that an Nd value of 1 did the trick pretty nicely. It was still a bit too shiny in spots, so I did use the image adjustment feature to clamp the roughness map by increasing the min value. While I never had much success doing this with a 16 bit file, the 32 bit EXR provided a lot of flexibility.



Then came the part that was always tough for me, the dialectic layer. I punched in the maps the same as above but it looked way off... THEN it hit me (you, I'm sure already know this, but, as I said, I'm not too bright)... Substance thinks of roughness as a continuum between shiny and not shiny. Maxwell thinks of roughness as a continuum between metallic and non metallic. Plugging the roughness map from Substance into the paint layer slot of my material was screwing everything up. Instead of mapping the roughness here, I left the roughness at 99 and added a coating at 1000nm thick. That seemed to do the trick for this model, but I think in the future, I may need to try some other options for more advanced roughness blending. The color was still a bit off from what I wanted, but honestly, I believe that is just because Maxwell did a more accurate job of rendering the effect light would have on it. I again used the image adjustments found in the Maxwell section to tweak the color a bit:



In the end I added a third layer with a very weak gloss coat and masked it a bunch to try and get more variation in the shine. This is the one area where I think I still need to hone my technique so as to accurately obtain a myriad of roughness values over a dialectic material:



I hope others have found this useful. More importantly, if anyone knows how to do this better than I've done it, please, PLEASE respond with a breakdown of your process!!!

[Mihai]

Substance thinks of roughness as a continuum between shiny and not shiny. Maxwell thinks of roughness as a continuum between metallic and non metallic. Plugging the roughness map from Substance into the paint layer slot of my material was screwing everything up. Instead of mapping the roughness here, I left the roughness at 99 and added a coating at 1000nm thick.

I wrote about this in another post, I don't remember exactly where but, when you have a situation of paint over a metal, you need three layers in Maxwell to mimick this:

Layer 3 (shinyness of plastic/paint) (set to additive mode!) -> use as layer weight the metallic map you get from Substance
-----------
BSDF set to a typical ND for shiny plastic/paint, something around 1.5
Roughness of this BSDF mapped with the roughness map from Substance

Layer 2 (main color of plastic/paint), use as layer weight the metallic map you get from Substance
-----------
BSDF set to 99 roughness
Use in the refl 0 any albedo maps you got from Substance

Layer 1 (the metal)
-----------
BSDF with metallic Nd and K. If you want to get something metallic looking I'd forget about the "IOR map" or anything like that from Substance. Just use an ND and K of what type of metal it's supposed to be. You will get a more accurate metal with Nd and K, especially if it's supposed to be a super shiny metal like Silver.

Now depending on how you did the material in Substance, you might have the same roughness map you used for the paint, or have a separate one just for the metal parts. Use whichever you want in the Roughness for this BSDF.

Same with the albedo map, you might have one made separately just for the metal, or just one overall for the whole material which you use already in the Refl 0 of Layer 2. Re-use it here in the refl 0 and 90 in that case.

Finally if you have scratches and bumps you can load that into the Global bump of the material.

So basically you have at the top a typical two-layer plastic approach, and both use a layer weight which is the metallic map from Substance to decide where the paint vs metal will show up.

You can also make the bottom Metal layer more complex if you want by duplicating the metal BSDF and have a slightly different roughness on it for example, and a bit less weight. Gives more interesting highlights.

[mjcherry]

I understand the concept in Maxwell, I just with there was an easier way to directly use substances. So now I need to figure out the best way to map the roughness off the gloss layer. I'll do that next.

I love Maxwell, but doing this in Arnold is soooo easy. But the results in maxwell (when I get get it to work) are better.

[Mihai]

There is a Python library for Substance called PySBS which:
"exposes all the graph editing capabilities of Substance Designer programmatic way allowing to create custom graphs"

It would be great to have the Maxwell devs create a few scripts that can take a Substance graph and automatically create the outputs and MXM based on a Substance graph, using both PySBS and the PyMaxwell libraries. These scripts could be run inside of Substance Designer...

[mjcherry]

OMG that would be amazing!

Take metals for example. Yes, if you want a nice new shiny metal the just using a preset in Maxwell is the way to go. But I actually have a few of these old movie lights (I'm actually a photographer/director who is learning 3D) and the metal is a rough, not very reflective mess. That's much easier to create in substance.

[mjcherry]

I'm hoping lebulb chimes in and shares his process as well.

[mjcherry]

There is a Python library for Substance called PySBS which:
"exposes all the graph editing capabilities of Substance Designer programmatic way allowing to create custom graphs"

It would be great to have the Maxwell devs create a few scripts that can take a Substance graph and automatically create the outputs and MXM based on a Substance graph, using both PySBS and the PyMaxwell libraries. These scripts could be run inside of Substance Designer...

How do you think we could go about making this happen? Would it have to be their developers or is this something I could hire a programmer to do?

[Mihai]

I think anyone with a decent knowledge of Python should be able to do it. I'm just guessing but I think that we may not even need to use the PySBS because that's ment to build graphs for you. What we need is to simply call a PyMaxwell script whenever you output the maps from Substance, and this script builds an MXM for you. Hopefully there is an option somewhere in the Substance settings to call a script when you choose to output the maps.

Ideally, there would be 3 scripts for specific substance cases: metals, metals + paint, plastics.

How does it work for Arnold? You mentioned it's easy to transfer materials to it.

[mjcherry]

Arnold handles materials differently even though it uses a Metal Rough workflow. In Arnold, roughness does not equate to metalness, there is a separate control for that. You simply plug in the maps generated by Substance, again as EXR files, using the Arnold 5 preset and plug in the maps. The material renders perfectly right off the bat.

I have been doing more and more rendering with Arnold and, due to the nature of my work, may find myself moving more and more in that direction, however, I would greatly prefer to stay in a Maxwell workflow. Maxwell is awesome for archviz and product and that does translate over to some of the work I do, "perfection" in materials is not what I'm usually going for. While I can certainly grunge things up using a traditional zbrush/photoshop/mari workflow with Maxwell, Photorealistic procedural rendering with Subtance is, I think, the future. I also have some projects coming up where I will have to work with some animals and groomable fur, so I think I'll be using Arnold for that, though I may try Shave and a Haircut if that still works with Maxwell v4.

Some examples of the kind of work I do:

[Mihai]

Nice work! So if I understand correctly you still have to create the Arnold materials yourself, it's just that there is no confusion as to what goes where.

I will try to grab the attention of Abraham at NL, maybe we can do something better than an output preset

I'm not sure why EXR maps would be needed in Maxwell. Maybe in Arnold it lets you have a linear workflow but Maxwell assumes maps have a gamma curve and so "de-gammas" the textures before rendering. Except when loading 32bit files, then it assumes they are linear. But either way it shouldn't make any difference when rendering. If you have some examples of using EXR textures vs PNG/JPG etc. in Maxwell and EXRs look better, please make a post about it.

[mjcherry]

I will have to do another test at some point. Maybe it's in my head. I know that for Arnold it definitely makes a difference and the literature on Substance seems to back that up. I don't think it has to do with Maxwell linerazing the file, but rather how much information is actually stored in the original map. I believe that Substance maps (internal) are in 32 bits. I could, however, be very wrong on this.