A silly idea to the developers

[oguzbir]

Hey,

I don't know if it could be done or ever discussed or not.
I must say that I know little about programming especially cuda..
Could it be possible to merge or mix the process of a directlighting and thus using the color or lighting info of the directlighting with the pathtracer ?

What I want to tell you is:
If we can mix the mesh preview kernel with 2 kernels in a final render process.
First we start render with directlighting than pause it and continue render with pathtracer.
By doing that pathtracer will multiply or add color values to the pixels ?

IMO, It might speed up the render process for large scenes?

I do not know if it makes sense or not. But I wanted to share the silly idea of mine.

Thank you.

[Daniel]

Can't you do that in post? It seems a little unnecessary to build a feature that takes a few extra seconds in another program.

[oguzbir]

Can't you do that in post? It seems a little unnecessary to build a feature that takes a few extra seconds in another program.

Actually what intended to offer is to have the option to mix the render kernels. Not rendering both kernels and mix them in post.
And not intending to ask for an option just to enhance the visual quality or something like that.
My offer is to speed up the render. That takes some hours to finalize the render, let's say 16000 samples.

If this works in programming, it might speed up the pathtracing or PMC renderer by creating a pre-calculated diffuse colour lighting information.
I might work like a pre calculated pass like it works on biased renderes. I might speed up the octane unbiased architecture.
The main reason is to speed up the pathtraced render with many light bounces.

not insisting for something I have less knowledge of.

just a maybe..

[MrFurious]

If I understand the OP's idea correctly to be able to mix render kernels would be invaluable addition for interior scenes. For example, I have an interior scene with a few small windows. Light floods in through the window holes but you have to increase the samples in pmc, path tracing and diffuse for the light to reach the furthers corners of the room, otherwise it's too dark and contrasty. Even with samples at maximum it takes forever to resolve cleanly. One could change the gamma setting but this effects EVERYTHING including the gamma of textures and picked colours and the image starts to look washed out.

However if we could combine path tracing with ambient, the ambient would act kindof like a fill light, brightening up the scene without effecting the gamma/tomemapping via imager. Would be a great way to control ambient illumination without having to place emitters everywhere.

[dsyee]

It sounds like you're asking for a mix of biased (DL/AO or similar) and unbiased (PT/PMC) kernels, which produces a biased render. Many (most?) other renderers do this already, or something similar. I don't think that's what OTOY has in mind for Octane, though...

[Edit] Remember "ambient lighting" doesn't really exist in real life...it's all indirect lighting, maybe with 1000s of bounces.

[MrFurious]

"ambient lighting" doesn't really exist in real life...it's all indirect lighting, maybe with 1000s of bounces.[/quote]

..and maybe we all move at 1000fps?

This isn't real life & we are producers of CGI- and that doesn't always mean we want 100% mathematically correct light transport. Even with 'Physically Correct' lighting you still need to tweak your scene with additional lights, tricks and whatnot to achieve a particular look or mood.

I mean, what IS real world lighting anyway? The clever chaps who develop the algorithms to produce real world lighting come up with a mathematical model.. a model. It's not real. It still takes an artists skill to interpret into something that resembles realism.

I'm a tad fed up with hearing 'not true lighting', or 'fake' lighting. At the end of the day your client couldn't care less if you used AO, PT, or HB (pencil) as long as you deliver a beautiful image. And I think Octanes implementation of AO is more than capable in many cases, and the speed increase is just what you need sometimes to get the job done.

Test renders - all AO no PS. Who gives a rats' if it's fake or not physically correct??

Anyway now that I'm done ranting, a combination of Ambient Occlusion, or even Simple Ambient (I haven't found a useful application for SI yet) and Radiosity would be immensely useful. There's a reason other render engines have it and that's because it gives the artist some flexibility in achieving the look they're after without having to rely purely on radiosity. As great as Octane is, I feel restricted sometimes when using DL, or Pathtracing. Sometimes those rays just don't go where you want them to, even if it is 'physically correct'. And adding more emitters just gives rise to more noise and longer render times to the point where in complex scenes, a biased render engine would be a better solution.

[dsyee]

You're preaching to the choir here. Nobody needs to be convinced that "faking it" can produce outstanding results. My point is that VRay, Mental Ray, 3Delight, Renderman, Corona, Redshift, and many others can all do that.

I just don't think that's the point of Octane. Octane claims to be "the world's fastest unbiased photorealistic renderer" (which to my knowledge it is), and depending on your needs that can have advantages and disadvantages. Simplicity and ease of setup are huge advantages in my opinion, and personally I'd rather OTOY not clutter things up with biased GI settings, AO maps, "fake" SSS, etc. and all the other stuff that's required for a "do-it-all" renderer. There are plenty of other renderers that were built from the ground up to handle those things. I'd prefer OTOY spend that time improving convergence speed and fleshing out features. Maybe they would disagree, but I hope not.

[ten]

Simplicity and ease of setup are huge advantages in my opinion, and personally I'd rather OTOY not clutter things up with biased GI settings, AO maps, "fake" SSS, etc. and all the other stuff that's required for a "do-it-all" renderer. There are plenty of other renderers that were built from the ground up to handle those things. I'd prefer OTOY spend that time improving convergence speed and fleshing out features. Maybe they would disagree, but I hope not.

Agreed. Whilst more options are always nice I love just throwing stuff into Octane and have it look pretty immediately. If I had time to arse around tweaking 1000s of settings id purchase vray. The GI settings tab in vray just makes me want to eat my own vomit.

ten

[MrFurious]

AO is Biased, and so is Diffuse. Octane claims to be the fastest Unbiased renderer yet it also offers AO and Diffuse and I'm glad they did because they are faster at the expense of accuracy, which trumps quality sometimes. Combine render kernels or multiple render kernels.. It's just offering more flexibility imho.

[daveyt]

In PMC you can change Max Rejects,

This can control the “bias” of the render. By reducing the value, the result will be more biased, but the render time will be shorter.