Will MILA material replace the MIA material?

#1

Is the MILA material only a mental ray layered shader or is it more than this?
Does it do everything the MIA material does (like Bump, AO, advanced refl and refr attributes…)?
Will it replace the MIA material in the future?

#2

I think it eventually will as it lends itself a lot better to brute force methods and ray efficiency in general. I can see a ton of benefits of the system compared to mia.

I haven’t tried and tested MILA enough myself to say it definitely does and will replace the mia in all situations. But I’m liking where it’s at, and the direction they’re heading with it.

The main benefit I see is that its a perfectly efficient layering shader system that is always energy conserving and will completely eliminate multiple light loops and shader calls(which can increase render times exponentially with traditional layering, where each shader in the layered group would have to be evaluated in full even if you’re only viewing 5% of one of them… with MILA this isn’t the case).

It is a very future proof system, they can keep adding MILA components per update and the same workflow remains.
It’s also a way of directing the user to have a more efficient set of shaders and not a ton of monolithic mias everywhere when the user probably doesn’t need all of mia’s features for every single material, cutting down shader evaluation time and rays needed.

As far as I know it has almost everything the mia library has.
It replaced the mia shader AO with a detail path tracing distance. So you can use very interpolated cheap FG for your whole scene and only brute force path trace crevices and the minute details on a per shader basis.
To me this is a great advancement of the same mia-AO/FG technique most mentalray users are used to.

One feature I’m really happy they kept is the single sample from environment which still works with mia_envblur, a huuuge time saver in evaluating glossys in certain scenarios. A feature I miss a lot in V-Ray.

A lot of interesting info here:
http://docs.autodesk.com/MENTALRAY/2015/ENU/mental-ray-help/files/shaders/layering/background.html

http://docs.autodesk.com/MENTALRAY/2015/ENU/mental-ray-help/files/shaders/layering/layering.html

#3

MR for maya was lacking a layered shader for ages, the most common way of bypassing this limitation was to use a technique combining MR materials, surface shaders and a layered texture node which was a very dirty solution. I am glad that we finally have a MR layered shader (this should have been done 13 years ago). That said, I don’t understand the timing for ditching the mia_material, I mean one thing at once. Users have spent years investing in the mia_material and now they are being said that it’s not the best shader in the house anymore, and are invited to drop it for a new one that also is a layered shader. Why not have a new version of the mia_material and a new efficient mr layered shader when needed, like other mainstream renderers?

#4

I don’t think they are ditching the mia at all. It will remain available I reckon for a long time just how the legacy dgs stuff is still there. My point maybe came across wrong, I’m predicting it will take over the mia in production primarily because of the speed advantages MILA has over mia in brutefroce scenarios. This isn’t due entirely to MILA being new code and mia being older code, it’s just the nature of monolithic shaders taking much longer to evaluate.

And since all rendering is heading in the direction of going full brute-force on everything MILA will eventually be the weapon of choice. Mentalray’s way of dealing with indirect illumination just needs to improve a little.

Use mia_x where you don’t need layered shaders, mila shaders where you do. A full MILA scene will be much cheaper for brute force scenarios, this is not to say hey you need to use mila or otherwise you can’t use brute force illumination methods, it’s just an advantage mentalray will have over other renderers which defaults the user to using monolothic shaders everywhere.

#5

All right man, thanks for all the info!

#6

i would say mila will take over mia materials in the future… but it will change slowly…
if you have a mila phenomenon its not so different to use but with all the new candies…

for example LPE or mila framebuffers… or the ability to create a shader lib that will work for mental ray and iray cause of MDL…

more info here…

http://blog.mentalray.com/2014/04/18/layering-in-maya2015/

http://elementalray.wordpress.com/2014/04/18/the-layering-library-mila-part-1/
http://elementalray.wordpress.com/2014/04/20/the-layering-library-mila-part-2/

#7

I’m planning on a mia to mila converter tool. Should make it one click to convert mia scenes to mila. I’ve already got code for mapping the glossy too, wrote some code to do it from mia to vray glossy a while back since they’re so different.

I’m a huge fan of the whole mila workflow and modular design methodology and am really looking forward to working with those shaders. Already made some phenomenons with nice mia-inspired AETemplates. I love the ability to create my own phenomenons and only expose the controls I want. Keeps me from using a giant catch all shader, like the mia, with piles of extra controls i simply don’t need for certain materials.

Also, I’m extremely happy about not having to create giant shading networks to fake a layered shader with a mia using masked texture inputs. I learned long ago layering even two mias means render time death as ray counts explode. Those days are gone! Thank god.

#8

that would be awesome! please let me know if you need some betatesting, i dont have maya 2015 just yet but I will get it soon… :beer:

#9

Hey, Any news on that script? :stuck_out_tongue:

#10

I would be forever grateful for such a tool on Maya 2015!

I find building out MILAs to be a huge pain, even though I love the ideas they represent. I don’t actually mind the big huge networks combining shaders and masking textures, I find them easier to think through and work with than MILA, but I hate the performance hit. If there were some kind of hybrid solution or at least a convienent way to convert between them, that’d be perfect!

#11

In fact, in 2016 Ext1 there is a tiny button in the render globals to convert mia_materials to MILA…, great, but it don’t do anything O_o

#12

Other than a couple specific built-in features/hacks that the MIA has that are handy in specific cases, I think the MILA is better in every way than the MIA. Though I do wish all the layer nodes it created could somehow optionally be auto-stacked in the hypershade to reflect their stacking order so I didn’t always have to do it manually.

You’re able to do so much more with the MILA using far fewer nodes than with the MIA + MISSS nodes.

What I find frustrating with the MILA is if you have 2 MILA shaders and then needed to blend them with a texture. There isn’t really an easy way to do that as far as I know without reconstructing a new MILA shader from the ground up with a million texture blend maps going into each attribute.

You can’t just layer 2 separate MILA’s without resorting to the same old ways of layering MR shaders together and suffering the render time hit.

If there was a way for MILA to have a new layer button called something like “shader layer” where it would let you pipe in the output of an existing MILA onto another MILA, but have it all be calculated as one giant single MILA with fast performance, that would be cool.

#13

My main concern is that MILA doesn’t have the builtin AO as mia_materials do. I use it in conjunction with colorbleed to fake GI without flickering nightmares when animating…

#14

It’s not IN the shader like usual, but it can be added as a pass… result is the same… 1 exr with beauty and AO… It does have less controls though…