[bograt]

Hi All,
I have a few questions regarding the nature of unbiased render engines, and what the future possibilities are.
Although I do not really know how the rendering engine works, I have had my theories based on the functionality of Maxwell render and Fryrender.
One of the biggest problems with these engines is how incredibly slow they can be, but I have always been so impressed by the render quality and how intuitive they are.
In my opinion one of the most impressive features of these engines is the multilight function and fryrender swap: http://vimeo.com/2406473
It seems clear that this is all possible because light is purely additive and white light is just a mixture of other wavelengths and it makes me wonder how much it is possible to pre-calculate.. e.g:
If I made a complex scene with various emitters and all surfaces diffuse grey with 50% roughness, based on the calculations made by the engine, in theory could I not change almost anything in the scene based on the calculated data?
Its fair to say that things like reflected/refracted caustics would be out of the question because there would be never be enough samples focussed in that direction to produce clear caustics but I cant see why with a high enough sample level far more than just lighting could be changed in post?
This came to mind again recently after I was trying to figure out how to create an animation of a room illuminated by a large video wall, I cant see why a whole animation could not be done in one render as it is only colour intensity and that is changing.
(I am still trying to work out if there is a way I can 'code' the video screen material to apply these changes in post)
This is something I have always been curious about so please share your thoughts!

Jules

[eric nixon]

This Swap thing doesnt have many uses, It would be useful for an interactive session with a client but they may get the impression everything can be done instantly - which would be bad.

Other than that it doesnt really work properly, because the image is no longer truthful, in the vast majotrity of real-world situations.

It looks very cool, but Random-control are known for hyping BS, remember realtime unbiased walkthroughs? so beware the hype. Maybe someones used this to good effect on the fry forum... - oh they dont have a forum anymore, nor an archive.. how odd.

[numerobis]

swap has been canceled and is no longer developed... the same applies to fryrender and the arion plugin support for everything else than max and xsi (...and the forum)

[polynurb]

there is something similar but it is more a standalone product:

http://www.motivacg.com/colimo_en.php

you need to have modo render/mental ray/thea render to produce the special passes but other than that it works with any 2d pixel input, so also with maxwell.

[Bubbaloo]

swap has been canceled and is no longer developed... the same applies to fryrender and the arion plugin support for everything else than max and xsi (...and the forum)

Seems they found found their next limit.

[bograt]

Ye for sure,
Random control keep changing their minds as cancelling products as their software is clearly full of potholes.
Eric, I agree with what you say about it being a bad idea considering how the client might take it and in all honesty I am very happy with the current system and how I see the future unfolding with developments in hardware,
But I am curious about how it all works. You say that it does not really work, I accept that currently there is no system or solution for this but is it really impossible to predict, calculate and simulate different materials and lighting post render?
e.g: (bearing in mind I do not really know how it works) One 'photon' fires out of an emitter, as it hits the first semi-rough surface it gets scattered slightly veering of by 10 degrees... the second photon fires... naturally the angle differs but this time it veers off by 5 degrees.. somewhere along the line they start to bounce off within 1 degree of the perfect angle of reflection. Assuming each 'photon' has been recorded would it not be possible to filter the photons according to the roughness and fresnel equations etc of the material to alter the properties of the material? So in the example I gave, all photons veering off by more than 10 degrees would be filtered more heavily to create a smoother material, altering the curves of the filter would effect the fresnel effects etc.... Yes it would get seriously noisy at even a slight change to the material but in theory is this not possible?
Now lets say you also record the angle at which the 'photon' leaves the emitter, could this not allow you to alter ies settings after rendering?
I guess it may take the same amount of time to render one frame with this function as it would to render every other possibility individually but I wonder what the limits actually are
I also accept that I may have completely misunderstood how Maxwell calculates...

[eric nixon]

I also dont know how it works, maybe motiva cg forum peeps know.

Try googling 'ColimoSSRMapGenerator'