MaterialX!

[vfxronin]

hey,

I just want to say that ILM is releasing MaterialX which is there shaders that been using StarWars the new hope and the new Rouge one.

http://www.materialx.org/

just wondering how dose the V-ray community feels about this news?

[^Lele^]

yawn.
and padding to meet the minimum characters for a post.

[vfxronin]

you are aware this is ILM shader, the very same shader that they use on movie that was rendered in V-ray!

yawn.
and padding to meet the minimum characters for a post.

[^Lele^]

That is NOT a shader.
It's a shader description language, which they used internally on their latest show, and now call (once again) open standard.
As long as you're Arnold, rMan, Maya and Katana centric, of course.

I'll put it as simply as i can: everyone speaks their own language, translation is done with dictionaries, complex as those may be, and for some "words" there simply isn't, and never will be, correspondence.
This is another language, like Esperanto, like USD, like Latin before them, and FBX.
It will become a Standard IF AND WHEN people will have converged on the set of functionalities this covers, and will have written all the required ecosystem of ancillary extensions and plugins.
Others can do more, TODAY, with different ways to go about it, and have done so for ages: I was porting shaders to and from max and lightwave, with custom writers and readers, fifteen years ago.
Nothing has changed, really, in the core methodologies, and i don't see it change any time soon: in the case above, i'd have to write in their language, with their syntax, and read from their language, with their syntax; and because neither of the apps was in the minds of the standard writers, i'd have to do so through a swathe of custom definitions.
The issue with it would be having to try and make it work for yourself, should it not work right off the word "go": gotta love "Open" standards.
And that can be a serious drain of resources, once multiplied by a number of different scenarios (for not only some things in some applications aren't 1:1 translatable, but they ain't even shaders or shader components to begin with. See displacement, to name one.)

But that's my take, i'd be all ears to see where this new tech is magical, in the details: what i read of the spec so far sounded grandly patronizing to me, aside from being, as i said, quite explicitly app-centric, and solving about zero issues of the ones still being show-stoppers for translations across apps, once one doesn't consider just four of them and calls it a day.

[jasonhuang1115]

Thanks for sharing your take Lele. From comments I read elsewhere that Anders Langlands (creator of alshader pack) shares similar concerns on how well each VFX house may take on MaterialX's spec when dealing with shaders/materials coming from multiple houses using different renderers.

Do you think wider adoption of MDL or OSL among renderers will be better or more practical than following the MaterialX spec?

-JH

[^Lele^]

Thanks for sharing your take Lele. From comments I read elsewhere that Anders Langlands (creator of alshader pack) shares similar concerns on how well each VFX house may take on MaterialX's spec when dealing with shaders/materials coming from multiple houses using different renderers.

Do you think wider adoption of MDL or OSL among renderers will be better or more practical than following the MaterialX spec?

-JH

Oh, glad to hear someone wiser than me read the specs and had the same doubts...
If i had to bet, but you should know i suck at that, yes, more than a simple descriptive schema, a proper shading language, with the ability to implement custom XXDFs, could potentially allow for a way out.
OSL is terribly, terribly limited still, and years have passed since its release, without much change in those shortcomings, while MDL seems a LOT more complete, closer to the metal, so to speak, and able to accept a much broader amount of customisation, if it seems to stop short of allowing custom BSDFs (but do not quote me on this. i need more play.).
Futher, it's all in NVIDIA's interest to play nice across the board of adopters, and proactively help them out, while the same, with all the good will in the world, cannot be said for Disney, Sony, and other VFX houses (see this case, or USD, or ABC, or OCIO. the list is nigh endless).

In general, i grew up in a world where "Standards" were created by non-commercial committees (EBU, IES, IEEE, and so on), and mistrust implicitly when a commercial entity releases their own tech, seemingly for free, and calls it a "Standard".
I always wonder where's the catch, and historically the catch was in tech which may have worked originally, but by the time it was implemented somewhere else, it was so broken, lowly performing, limited, and convoluted, that just trying to make it work wasted huge resources, for the sole benefit of whoever else will later get the tech advancements, should one choose to share them.

Views differ on what's to be done going forward, and i'm sure well biased in mine.
Yet, for the humongous amount of work it'd be, i'd MUCH prefer if Chaos went its own merry way, without the presumption of calling them standards for everyone else, and make the most of the tech we already have, as far as pipeline and cross-app talk is concerned.
The central development will ensure that changes are to the benefits of all users, while the SDK allows for huge latitude for those which want more customisation.
After all, V-Ray's in so many apps now, that all it needs is a vrscene to be pointed at, and it'll render it as it was saved.
It may need (potentially a lot) more polish work, but the direction's quite clear, and it works TODAY.
Not perhaps in a year or three, perhaps.
Perhaps.

[jasonhuang1115]

After all, V-Ray's in so many apps now, that all it needs is a vrscene to be pointed at, and it'll render it as it was saved.

Thanks for sharing your views Lele. I can't help but highlight the above...

[t.petrov]

Do you think wider adoption of MDL or OSL among renderers will be better or more practical than following the MaterialX spec?

You're generally missing the point or you're mis-interpreting the spec.

The main goal of MaterialX is to solve the problem of texture assignments. This is the main problem they are solving!
Also the spec defines some texture compositing nodes and if they are used and supported by all parties (renderers, dcc apps, ect) then sharing materials will work fine.

The interesting part starts when there are custom shaders involved. It is not clear yet how this will work and it is up to the dcc vendors to do a good job and make our life easy.
If they do it, then MaterialX will be a success and it will work. Otherwise it will fail and it will be used only by ILM/Lucasfilms.

Also the MaterialX spec doesn't require any particular shader implementation. It is very generic and makes it possible to mix and match osl, mdl, glsl, c/c++ native shaders, etc.

[jasonhuang1115]

Happy to learn from you comments, Petrov. Thanks!

[oglu]

its now opensource and the docs are online...
http://www.materialx.org/

[oglu]

is there interest from chaosgroup to support materialX..?
im still ab bit confused how this should work in the end... the pdf isnt clear to me if we would need native vray materialx shaders or does vray convert matX shader to vray materials.?

[vlado]

I'm probably even more confused than you I'm still pondering in what way, if any, we should attempt to support it.

It was all fine and clear when the idea was to have some common shading graph description format. But then they added other things to the specification like material assignments, light/shadow linking, primary/secondary visibility, and trace sets.

So right now, I have no idea what it is and how it would fit into artists' workflow exactly...

Best regards,
Vlado

[pixelmonk]

J Cube has already created "Multiverse Studio" to handle MatX transfer with Maya, Katana and Arnold. It appears they already landed a deal with Foundry for distribution:

https://vimeo.com/225919845

https://www.foundry.com/news-awards/...e-announcement

[oglu]

the question still remains how do you render those materials in vray...

[vlado]

J Cube has already created "Multiverse Studio" to handle MatX transfer with Maya, Katana and Arnold. It appears they already landed a deal with Foundry for distribution:

https://vimeo.com/225919845

https://www.foundry.com/news-awards/...e-announcement

Yes, I know. However we don't plan to support Multiverse ourselves.

Best regards,
Vlado