Isotropix Forums

Denoising of individual AOV´s

General Discussion about Isotropix and CG related topics

Denoising of individual AOV´s

Unread postby machieb » Tue Nov 02, 2021 6:10 pm

Hello,

is there any way to denoise individual AOVs in Clarisse? And/or is such a feature planned in Clarisse´s nextgen renderer Angie?
In many other renderers like Arnold, Vray, Redshift, Renderman denoising individual AOVs is possible.

Greetings Marcus
machieb
 
Posts: 66
Joined: Fri Nov 29, 2019 11:20 am

Re: Denoising of individual AOV´s

Unread postby machieb » Tue Nov 09, 2021 10:49 pm

Is there really no way in Clarisse to denoise individual AOVs?
machieb
 
Posts: 66
Joined: Fri Nov 29, 2019 11:20 am

Re: Denoising of individual AOV´s

Unread postby sam » Tue Nov 09, 2021 10:59 pm

Hi there,

Not directly and the Optix denoiser is only trained on beauty. It can work on per light beauty I guess but you need to make a setup in Clarisse to denoise individual AOVs using layers.
Sam Assadian
Isotropix
CEO/Founder
User avatar
sam
 
Posts: 1719
Joined: Fri Jan 25, 2013 11:33 pm

Re: Denoising of individual AOV´s

Unread postby machieb » Wed Nov 10, 2021 4:12 pm

Hello Sam,

thanks for your answer! In that case the optix denoiser for clarisse is not useable for us at Trixter/Cinecite. We render many AOVs for compositing. It would take too much time to setup AOVs on layers and the storage space would also be a problem. I hope an advanced denoising system/pipeline/workflow for AOVs in clarisse is planned for the future?!

Greetings Marcus
machieb
 
Posts: 66
Joined: Fri Nov 29, 2019 11:20 am

Re: Denoising of individual AOV´s

Unread postby sam » Thu Nov 11, 2021 7:02 pm

Hi Marcus,

I don't think denoising is actually part of the rendering. It only is when it is actively participating to adaptive sampling during rendering which is only the case for prman (pixar internal version I think)? Not even sure. Well, when that's the case, the idea is that the renderer progressively renders. At some point in time, the denoiser kicks in in parallel and resolve the variance to let the renderer know if it should stop or not. That denoising process can takes up to several minute AFAIK each time it kicks in.

So most of the time, denoising is a post process and as such, it is an image filtering.

All that to say that unless you have an active and super reliable denoiser or denoising pipeline that actively drives a current render, I think that the best options are third parties solutions such as OptiX (which is real time), Intel Open Source Denoiser, AMD, Neat etc...

I'm not even mentioning the big issue of temporal denoising consistency...

So in Clarisse, the denoiser is really here to improve quality of the display feedback and to let the user use it on final render when it makes sense because the results are good.

Right now, I personally don't know any denoisers that would offer, what I consider, predictable and reliable results...

Will we bring other denoising solutions to Clarisse? Yes if they tick all our boxes: real time and efficient. Otherwise denoising has more it's place in a compositing application since it's a post process rather than in a interactive/RT renderer.

Am I making sense?
Sam Assadian
Isotropix
CEO/Founder
User avatar
sam
 
Posts: 1719
Joined: Fri Jan 25, 2013 11:33 pm

Re: Denoising of individual AOV´s

Unread postby machieb » Fri Nov 12, 2021 12:47 pm

Hello Sam,

thank you for your detailed explanation!
It makes total sense what you write.
Denoising is like you said more a post process and so a task of the compositing application.
So the compositing application should offer posibilities to denoise individual AOVs.
I´m very excited about Clarisses Next Gen Renderer Angie and wish you the best for the completion and the release.

Thanks Marcus
machieb
 
Posts: 66
Joined: Fri Nov 29, 2019 11:20 am


Return to General Discussion