ILM's Material X and Nvidia's MDL

[Morne]

Hi Guys

Anybody have any thoughts on these and which you think will prevail in the long run?

So far we have V-Ray support for MDL, not sure about Material X

For those that don't know what it is:
http://www.materialx.org/
https://www.nvidia.com/en-us/design-...tion-language/

[jasonhuang1115]

I am not an expert...but time will tell...for portable material descriptions, USD and J-Cube's Multiverse approach is also in the play....or maybe something from Chaos Group will prevail...

[oglu]

Over a year later its in a better shape.
But still not final.

https://github.com/materialx/MaterialX

https://github.com/materialx/Materia...rGeneration.md

and the Arnold shader whitepaper.
https://github.com/Autodesk/standard-surface

[gandhics]

Any "universal" material is a day dream.
Or.. you need to accept the fact that you will use only least common denominator of all renderers.

At least "universal" map tree could be possible with OSL.
Material? Good luck.

[^Lele^]

Any "universal" material is a day dream.
Or.. you need to accept the fact that you will use only least common denominator of all renderers.

this: https://www.youtube.com/watch?v=oXJvOJNUlCs

[oglu]

Bringing a complex shading network from Max to Houdini or Katana could be helfpfull i would say.
And a basic material that would work everywhere is also nice to have for background stuff.

[^Lele^]

Bringing a complex shading network from Max to Houdini or Katana could be helfpfull i would say.
And a basic material that would work everywhere is also nice to have for background stuff.

I agree on the hope.
In my own experience, however, while they all try and do much the same thing, they all get there differently.

In other words, writing in a file that your reflection is of GGX type, white, and with a gloss of 0.8 (or in PBR lingo, roughness of 0.2) ensures no consistency in the results when translated across engines.
Some will have a minimum glossiness, others will have slightly different microfacet masking model, the glossiness amounts won't be exactly linear, and mean a different thing to each, and so on and so forth.
Heck, even on diffuse models things do not match...

Then we have shading networks, and DCC apps with a set of tools which won't map (and thank goodness for variety, i say!) exactly, or at all between them.
So, we either agree on a subset of functionalities, the minimum common denominator Changsoo was mentioning, or we expand everyone's capabilities to match a super standard; but why would someone give away IPs to competitors?
And if they did, and everyone rendered the same images, with the same performance, what would the differentiator be?
That, if those IPs and algorithms even worked within the engines' different code bases, which they often do not.

These formats will surely allow you to recompose most - feature-mappable - shaders correctly, but the ability to obtain identical results isn't with the formats, but with the engines which interpret the data to render the image out.
And that part of the road is very long, that i can see.
Then again, i am often wrong.

[savage309]

https://xkcd.com/927/

Best,
Blago.

[^Lele^]

https://xkcd.com/927/

Best,
Blago.

you read my thoughts.

[oglu]

I dont see it that way. I dont need the exact same image. I dont expect to get the same result rendering in vray or arnold.
But i dont want to reconnect 500 shader with 30 nodes each time i need to go to an other tool. It not about the shader its about the hole shading graph and textures.

[Morne]

....
But i dont want to reconnect 500 shader with 30 nodes each time i need to go to an other tool. It not about the shader its about the hole shading graph and textures.

+1000 this

[^Lele^]

I dont see it that way. I dont need the exact same image. I dont expect to get the same result rendering in vray or arnold.
But i dont want to reconnect 500 shader with 30 nodes each time i need to go to an other tool. It not about the shader its about the hole shading graph and textures.

And why would you need MDL or MaterialX to do that?
Any format which you create a reader for can accomplish the simple task, and fail just as equally where there is no feature parity to leverage, no equivalent dictionary entry, forget shading models.
There is nothing radically new in either, it's all been done before in-house a million times, without neither, for the best part of two decades, with very similar limitations and results.
None of those in-house formats ever claimed to be "universal", i'll give you that. ^^

[oglu]

Yes inhouse and thats the issue. We work with a lot of other teams. We need rules to work together. And get scenes and shots from one studio to an other.

[^Lele^]

I personally don't think delegating to third parties is either less work, easier work, more comprehensive work, or the resolving kind of work.
I call it sweeping under someone else's rug.
But of course we all have our opinions on what a pipeline should look like.

[oglu]

You are maybe right. But cause this is an Autodesk project we have this native in maya. If we need data from an other studio its easier than developing our own solution.