This forum uses cookies
This forum makes use of cookies to store your login information if you are registered, and your last visit if you are not. Cookies are small text documents stored on your computer; the cookies set by this forum can only be used on this website and pose no security risk. Cookies on this forum also track the specific topics you have read and when you last read them. Please confirm whether you accept or reject these cookies being set.

A cookie will be stored in your browser regardless of choice to prevent you being asked this question again. You will be able to change your cookie settings at any time using the link in the footer.

Another "QTMGC" mess 🥴
#21
To me vsnlm_ispc, vsnlm_cuda and KNLMeansCL all seem to produce the same output.
I'm rewriting a bunch of stuff there, to make things more user controlled, so you might want to try different settings under "Filtering->Vapoursynth->Misc->Prefer", since I will probably add more options there.

Cu Selur
----
Dev versions are in the 'experimental'-folder of my GoogleDrive, which is linked on the download page.
Reply
#22
...update to KNLmeans ..

I've just ran a test using KNLmeans on a video clip .. And the encode process speed wos like a astounding 80+ fps while using x264 preset very slow + heavy denoiser!? 
This can't be right.  I remember vividly the process took way longer to complete using a heavy denoiser like knlmeans (using very same hardware btw)..  Even while using a gpu acceleration !?

Hence, this most certainly explains why the quality is so terrible (Quality VS Quantity/Speed) !

(18.07.2026, 19:02)Selur Wrote: so you might want to try different settings under "Filtering->Vapoursynth->Misc->Prefer", since I will probably add more options there.

Cu Selur


Do you actual mean "Filtering->Vapoursynth->Misc->Preview"?

Or the "prefer" settings under  "Filtering->Vapoursynth->Misc->Source" ?


I don't see  "Filtering->Vapoursynth->Misc->Prefer" tab anywhere you see !
Reply
#23
Sorry, meant: "Filtering->Vapoursynth->Misc->Script->Prefer"
Just uploading a new dev (should be up in ~10min), which adds some additional controls there.
[Image: grafik.png]

Cu Selur
----
Dev versions are in the 'experimental'-folder of my GoogleDrive, which is linked on the download page.
Reply
#24
ah, forgot to update the scripts, will upload a new version in ~20min => upload is finished (last version for today)

Cu Selur
----
Dev versions are in the 'experimental'-folder of my GoogleDrive, which is linked on the download page.
Reply
#25
(18.07.2026, 20:53)Selur Wrote: ah, forgot to update the scripts, will upload a new version in ~20min => upload is finished (last version for today)

Cu Selur


Thanks, i will download and have a look  Rolleyes


 Update:  I have installed latest dev and checked again.  Nothing has changed at all about knlmeans output quality.
A couple things i have noticed about that filter, is that i needs much MUCH higher strength values to notice any difference visualy !  And it does a bad job flatten / smoothen the image while retaining details at the same time compared to what it used to be  Confused !!

That's the first thing that i emediatly noticed when using that denoiser again after a long time...

Changing the "prefer" settings under "Vapoursynth → Misch → script → prefer" doesn't improve anything at all ! 


Maybe it's me , but i'll do an seperate install of an much older build of hybrid and see what the end results are using knlmeans with that build..

thanks anyway,*
Reply
#26
If you find a real difference let me know how the Vapoursynth Scripts differ. Smile

Cu Selur
----
Dev versions are in the 'experimental'-folder of my GoogleDrive, which is linked on the download page.
Reply
#27
(Yesterday, 06:00)Selur Wrote: If you find a real difference let me know how the Vapoursynth Scripts differ. Smile

Cu Selur

I have installed Hybrid rev 2024.06.16.1 old geezer  Big Grin , to see if output quality KNLmeans denoiser differs from latest dev releases.
And alas for reasons unknown to me, i can't see any difference  Huh  

Maybe, iam so spoiled to see the massive visual improvement of using A.I / ML Models in combination with denoisers such as knlmeans to upscale / restore low res media files over the years. 
That I am no longer impressed by the results of KNLmeans as a seperate filter Tongue .

Is it possibruu , we are that far gone already 😐 ?
Reply
#28
Okay, so nothing to fix. Big Grin

Cu Selur
----
Dev versions are in the 'experimental'-folder of my GoogleDrive, which is linked on the download page.
Reply
#29
(Today, 05:09)Selur Wrote: Okay, so nothing to fix. Big Grin

Cu Selur

...erm..maybe.. Dodgy ..

I can't put my finger on it.. But iam defenitely sure something is off with that particular denoise filter KNLmeans... 
Is it because of many windoz overhauls, updates..patches etc.. Who knows ¯\_(ツ)_/¯ 

But one thing is certain, the strength values need to be much MUCH higher all of a sudden to see an significant change !?
Like way in the past strength values like 2,00 wos way to much ..  Now, values like 3,0 pales by what it used to be🤔💭


Maybe, ill dig up some of my old media files i have used back in the day to test it with this filter for comparison.


regards,
Reply
#30
The normal denoiser shouldn't have changed at all.
It's the version from:
https://github.com/Jaded-Encoding-Thauma...knlmeanscl
and there hasn't been any real changes (aside from automatic building and distribution) for years.

Cu Selur
----
Dev versions are in the 'experimental'-folder of my GoogleDrive, which is linked on the download page.
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)