OTOY Forums

Hi ,

I am using 2024.1 R2/R3 c4d octane plugin and there is a visible alpha edge on Denoised Beauty. In older versions, this was not happening. I am saving in Octane Exr DWAB.
Any tip on what changed? Do I need to switch something on that I am not aware of? In older version 2023.1.2, the same settings, give correct results in both images.

Thank you

Miki

Are both associated alpha encoded (a.k.a. the equivocal misnomer "premultiplied" term)?

This Octane "Alpha" page is a preliminary write up on the topic, if it is of any help.

Hi,

it doesn't matter if the Premultiplied alpha is set to "none" or "encoded", in both cases, the final alpha edge is darker in Denoised Beauty. This was not happening with older versions of octane. With default settings, Beauty and Denoised Beauty had exactly the same alpha on default settings.

I tried multilayered file save and single layers. In both cases, alpha in Denoised Beauty is handled differently as in Beauty.

Then I tried an older octane version and on default settings both images were identical. That means something is different in the newest Octane version. And I can't use Camera Imager/Denoiser till this bug is solved.

Thanks for your effort though.

Best

Miki

Hi Miki,

Dino was able to reproduce your issue, and is related to core SDK, not plugin.
The devs will look into it for next SDK update.

ciao,
Beppe

nemcekmiki wrote: (...) Premultiplied alpha is set to "none" or "encoded", in both cases, the final alpha edge is darker in Denoised Beauty. (...)
With default settings, Beauty and Denoised Beauty had exactly the same alpha on default settings.
I tried multilayered file save and single layers. In both cases, alpha in Denoised Beauty is handled differently as in Beauty.
Then I tried an older octane version and on default settings both images were identical.
That means something is different in the newest Octane version.

The denoising factor was indeed of high suspicion of causality, but as you might know, trouble-solving is a series of basic questions to ones that may be considered st*pid, which are not. This is to rule-out with more certitude certain elements leading to the "culprit". This reply provided more information than expected, thank you.

Hey Guys,

thank you for looking into this. Yes Dino reacted on FB group where I posted the problem too and let me know its caught up.
Very happy to be helpful. Lookig forward for the next version.

BTW: since there are two types of denoise engines now, the second one "Open Image Denoise" doesn't do this issue, only octane denoiser. However, I don't know what is the difference between these denoisers.

Best

Miki

nemcekmiki wrote: BTW: since there are two types of denoise engines now, the second one "Open Image Denoise" doesn't do this issue, only octane denoiser. However, I don't know what is the difference between these denoisers.

Octane's own "Spectral Denoiser" is an in-house developed denoiser specifically for Octane's spectral core engine.
Open Image Denoiser is Intel's one, as its name suggests, is open thus allowing developers to support it widely.

The particularity with the OID in the compositor is the possibility for instantaneous ("real-time") denoising.
Lino has it setup as such as shown in one of his posts (perhaps on FB), it is also demonstrated here where the AOV Group Output is the viewed image buffer, instead of the default Octane's Imager (Octane C4D shown here).

john_otoy wrote:OctaneRender 2024.1.2 has been released: viewtopic.php?f=24&t=84359

mojave wrote: Changes since 2024.1.1

Fixes

• Fixed denoised render AOVs and volume mask render AOV containing an unintended alpha premultiplication when kernel alpha channel is enabled.

Thank you so much for your reply!

Glad it has been fixed with the new version.

Best

Miki

I’m currently getting this same issue with some files in macOS Sequoia 15.3, C4D 2024.4.1 (and tested in 2025) + Octane Studio 2024.1.2

Also described and said to be resolved here:
viewtopic.php?f=30&t=84444

Files which have rendered perfectly for me with transparency previously are now having issues with 2024(R3) and latest plugin.

Can any devs check or verify this has been updated and is fixed in the latest macOS (Metal) build 2024.1.2 ?

Many thanks!