How to enable GPU for MCTemporaldenoise filter?

[Selur]

I am attaching the logfile.

sadly that didn't work.
Also does the Vapoursynth Preview work for you?

[Selur]

Think I found the problem, depending on the resolution CTMF is used, which Hybrid doesn't load.
-> looking into it

[Toothache]

Think I found the problem, depending on the resolution CTMF is used, which Hybrid doesn't load.
-> looking into it

I'll try attaching the logfile again anyway, in case you might need it.

[Selur]

Will send you a link to a new dev version for testing in a few minutes via PM.

Cu Selur

[Toothache]

Hey, did you forget to send me the dev version?

Earlier you recommended SMDegrain - could you tell me what adjusted settings you would use, for my use case? (Removing heavy grain from an HD source that is otherwise excellent quality.)

BTW, some of the settings appear greyed out in SMdegrain - the prefiltering dropdown menu, for instance. Hybrid 2019.09.30.1

[Selur]

Thought I send you a link.
Looked at the GUI and seems like I accidentally copied some of the Avisynth controls over to Vapoursynth.

-> send you a new link to a dev version via pm

Cu Selur

[Toothache]

Yes, Temporaldegrain is now working fine, and eliminating grain quite well with the default settings. I'll run some more tests to see if everything is OK.

Let me point out a couple of bugs in KNLmeansCL:

1) Selecting any channel other than "auto" or Y gives an error. If I select YUV or UV, the encoding crashes with a message that the filesize is too small. I'll send the debug log if you need, although you could probably replicate it yourself.

2) Selecting any weighing function other than the default (Welsch) seems to cause no effect on the output file. That is, I see the GPU working, the encoding finishes with no errors, but the output file doesn't get denoised at all, and visually seems identical to the input file. Even if I crank up the denosing strength.

[Selur]

2) Selecting any weighing function other than the default (Welsch) seems to cause no effect on the output file. That is, I see the GPU working, the encoding finishes with no errors, but the output file doesn't get denoised at all, and visually seems identical to the input file. Even if I crank up the denosing strength.

I can reproduce this (only tried one sample), from the looks of it Hybrid properly sets wmode, so this a bug in the filter or some of the other parameters need to be adjusted for those modes to work properly.
Best complain to the author of the filter over at https://github.com/Khanattila/KNLMeansCL/issues (or over at doom9s), doesn't seem like a bug in Hybrid.

1) Selecting any channel other than "auto" or Y gives an error. If I select YUV or UV, the encoding crashes with a message that the filesize is too small. I'll send the debug log if you need, although you could probably replicate it yourself.

Opened a YUV (Yv12) source and tried Y/UV/YUV and got no error.
(Note that for high bit depth, color is always converted to RGB30
Selecting 'RGB' throws an error, same will probably happen when the source is RGB and not YUV and you select a YUV color mode.
I added a workaround for this by adding an additional color conversions.
-> send you a new link to a dev version for testing which does these conversions.

Cu Selur

[Toothache]

1) Selecting any channel other than "auto" or Y gives an error. If I select YUV or UV, the encoding crashes with a message that the filesize is too small. I'll send the debug log if you need, although you could probably replicate it yourself.

Opened a YUV (Yv12) source and tried Y/UV/YUV and got no error.
(Note that for high bit depth, color is always converted to RGB30
Selecting 'RGB' throws an error, same will probably happen when the source is RGB and not YUV and you select a YUV color mode.
I added a workaround for this by adding an additional color conversions.
-> send you a new link to a dev version for testing which does these conversions.

Cu Selur

My input file is an 8 bit x264 YUV, not RGB. I'm encoding it to 10 bit x265, is that what you meant by high bit depth?

[Selur]

My input file is an 8 bit x264 YUV, not RGB. I'm encoding it to 10 bit x265, is that what you meant by high bit depth?

The important question is what color space and bit depth the content is in the filter chain when fed to the filter.

Cu Selur