mental ray diffuse pass Problem.

[Maximus]

Hi, i found a (bug?) problem with mr passes.

basically MR doesnt take into account the photographic exposure into the diffuse pass.

sphere on a plane > physical light or portal light as illumination
plug a photoexposure, add a diffuse pass.
you see the main (beauty) pass with correct intensity, if you switch to the diffuse pass everything is bown out.

so how you export a diffuse pass using portal/physical?

edit: actually it does the same problem with all lights, not only portal/physical. Apparently the diffuse pass doesnt take into account the photographic exposure.

[Kzin]

the lens shader does not work with framebuffers.
best way is to render without exposure and do the tonemapping in post.
otherwise, if its only the diffuse pass you can create a second pass with lens sahder and renderout only the diffuse.

[Maximus]

Well i just found out that the lens shader doesnt work with any of the passes, is this a Softimage problem or Mental ray? kinda surprised about this actually.

Thanks for answer Kzin

[Saturn]

framebuffer should not have any tonemapping applied otherwise you will not able to recombine them to make beauty render.
The maths simply doesn't work anymore.

[Maximus]

ok but i'm a bit confused here. So why even put sun+sky+portal + photoexposure whole system if you cant eport it in passes? That makes no sense.
Imo this is kinda stupid. you should let passes produce the same output as the beauty you created. Not be forced by the software that doesnt allow you to.
Infact i think its only MR that behaves this way.

thanks for the answers.

[Kzin]

ok but i'm a bit confused here. So why even put sun+sky+portal + photoexposure whole system if you cant eport it in passes? That makes no sense.
Imo this is kinda stupid. you should let passes produce the same output as the beauty you created. Not be forced by the software that doesnt allow you to.
Infact i think its only MR that behaves this way.

thanks for the answers.

because it was a user request. users which thoughts it would make more sense to render out to jpg instead of output the image(s) for compositing.
it was implemented in maya like this and users wants the same thing in xsi.
so its not the fault of mr, its the fault of users which dont know what they do, or lets say, they dont know how things work and all has to work like they want.

[Maximus]

ok but i'm a bit confused here. So why even put sun+sky+portal + photoexposure whole system if you cant eport it in passes? That makes no sense.
Imo this is kinda stupid. you should let passes produce the same output as the beauty you created. Not be forced by the software that doesnt allow you to.
Infact i think its only MR that behaves this way.

thanks for the answers.

because it was a user request. users which thoughts it would make more sense to render out to jpg instead of output the image(s) for compositing.
it was implemented in maya like this and users wants the same thing in xsi.
so its not the fault of mr, its the fault of users which dont know what they do, or lets say, they dont know how things work and all has to work like they want.

Yeah but not everyone and not always you want to send out to compositing every render you do, sometimes are simple things which doesnt require a whole lot of compositing, and sometimes you might need a single pass or diffuse pass to retouch something in those simple images.
So again i found pretty limited not giving users the ability to do that, actually there is no excuse at all, since you wont take away the "correct way" to do compositing with the correct math. you just add something.

Anyway no point arguing this anymore. I wonder why other engines have it, they are all stupid then.

[Saturn]

Anyway no point arguing this anymore. I wonder why other engines have it, they are all stupid then.

This have nothing to do with mr, this is a CG or more a maths logic. Every rendered are constrained to that.

Adding 2 pass that have been tonemapped IS and will give you a DIFFERENT result than adding 2 passes and tonemap it after.


The correct way is to not tonemap your beauty render at all if you use framebuffers, do it in comp.
OR
Don't use framebuffer

[Kzin]

Anyway no point arguing this anymore. I wonder why other engines have it, they are all stupid then.

which others have it and render their framebuffers tonemapped, means render them wrong?

[Maximus]

you just have the ability to render out passes that matches the beauty pass.
So for example if you tonemapped your beauty pass with a saturation +20, and you export a diffuse pass, you are able to get that saturation +20 embed in the diffuse pass.
This is more than logic to me.
Despite its correct or less, what an artist does with compositing or rendering shouldnt be chained to what the engine itself allow you, to mention one Vray let you do this, because the vray camera (which in this case is the photoexposure even if its ridicolous to compare the two since the xposure is just crap) add to the final image the variations you want (call it tonemapping) and when you render out the beauty pass, those variations are part of the beauty pass like it should be.

Now if you want to add render channels and passes, they are calculated based on the beauty pass wich has the camera added to them. So your beauty/reflections/whatever pass will have the same exact values to re-create the beauty pass you had before. This is not possible with mental ray, mental ray gives you the ability to create a beauty pass with all the modifies you want, but the xposure dont affect the passes. Which again it can be exact, math exact like you want, still its completely dumb and not artist oriented. Since i dont need to be a math/comp guru sometimes with simple renders, and i'd like some flexibility.

[Kzin]

you just have the ability to render out passes that matches the beauty pass.
So for example if you tonemapped your beauty pass with a saturation +20, and you export a diffuse pass, you are able to get that saturation +20 embed in the diffuse pass.
This is more than logic to me.
Despite its correct or less, what an artist does with compositing or rendering shouldnt be chained to what the engine itself allow you, to mention one Vray let you do this, because the vray camera (which in this case is the photoexposure even if its ridicolous to compare the two since the xposure is just crap) add to the final image the variations you want (call it tonemapping) and when you render out the beauty pass, those variations are part of the beauty pass like it should be.

Now if you want to add render channels and passes, they are calculated based on the beauty pass wich has the camera added to them. So your beauty/reflections/whatever pass will have the same exact values to re-create the beauty pass you had before. This is not possible with mental ray, mental ray gives you the ability to create a beauty pass with all the modifies you want, but the xposure dont affect the passes. Which again it can be exact, math exact like you want, still its completely dumb and not artist oriented. Since i dont need to be a math/comp guru sometimes with simple renders, and i'd like some flexibility.

use vray then. its easy as that. dont know what i should say other then that.
i could now write again that you should not tonemap your image to correct the lighting, i am pretty sure its what you want. if you work in the correct way you dont need tonemapping per ray. simple color corrections like you want are better doing in post, more control and faster feedback (besides that you can do this on a shader basis).

edit: if you want such a feature in mr, then open a thread in the mi forum, discribe what you want and why you need it. you will get awnsers there and perhaps there are some more user which needs it.

[Saturn]

"Since i dont need to be a math/comp guru sometimes with simple renders, and i'd like some flexibility."

Framebuffer maths are just ADD, SUB, MULT,DIV nothing to do with being a maths guru
Framebuffer purpose is to recompose the beauty render by adding all framebuffer to each others.


But again if you tonemap meaning you have a gamma (eg. LIN to sRGB ) or a contrast ( eg. Burn highligths ).
You can't ADD the framebuffers all together if they are tonemapped. Have a try in a compositing package and emulate the behaviour you will see that doesn't give you the result.
They are only a few action that will work, ( eg. Exposure, white balence,) because they are simple operation ( + , - , * , / )
,

Vray doesn't do any tonemapping at the lens level, nothing prevent you to do the same in mr too !
Remove the photographic exposure, and use the color managment of xsi to display the result.
Your render will be in Linear and not tonemapped and you will have more flexibility to do what ever you ever dreamed to do !
It is not that hard !