"THE PATCH" - for GH2 Hack - Developed by APEFOS

frullaccia

frullaccia

@apefos Sete by Apefos. No, something is wrong. I cannot read your file...

producer

This is the very first time I'm satisfied with GH2 because of your "ISO8000" patch, APEFOS! THANKS!!!

A simple test with ISO 6400 - 8000:

zsound

@Producer Your grading is fantactic, i would like to know where can i find the FilmVision Pro plug in, it's hard to find information on google..

Thanks

apefos

iso 8000 and below versions, same as previous upload, but with fallback for the 24p:

frullaccia

@apefos You should zip the iso8000 and sete (PAL) version. Thanks!

frullaccia

@apefos If you don't mind, this is the Apefos experimental patch for Pal users (see previous page): sete.

frullaccia

@apefos These are the b and j settings by Apefos, ISO 8000 and below versions.

producer

@zsound: My custom made FilmVision Pro plug-in is still not released. Since more than 2 years ago I started developing it and although I got some really nice pro results with the plug-in applied on videos taken with various cameras/video cameras, I could never reach something "real" with anything shot with Panasonic GH cameras. The results were nice at all, but not like what I expected to be... until the greatest patches by APEFOS.

Now I can strongly recommend the latest Apefos patches to every GH2 owner. I can surely say that everything created before (IV2, Moon T's, etc.) is too many steps behind Apefos "The Patch"- series. Sorry to say, but I have my deep doubts that latecoming Moon T9 can do something similar, especially if bitrate and some other non-sense details are its target as usually... I think that every next patch must be considered with the achievements of Apefos.

apefos

@frullaccia thanks for the uploads

@producer thanks, and nice images!

the J and B, 8000H and 8000S, inside the 4versions upload does not have fallback in 24p.

the new uploaded J and B 8000H and 8000S have fallback in 24p.

This can make difference in noise/textures in iso 8000 - 6400 in 24p 1080p 24H and 24L.

Please compare prevoius B with new B, and previous J with new J to perceive which ones shows better noise/texture. (it does not make so much sense to compare B to J because datarates are different)

Thanks

producer

@apefos: Would you explain with simple words about what you call a "fallback"? I already read what you wrote, but not sure I understood it right. Thanks!

Manicd

I see you use setG matrix for 24p fallback? Does this mean it can be good at shooting iso 160-8000?

apefos

@producer the fallback is a second matrix with a higher quantization. when the first (main) matrix cannot handle the noise or a motion due to low quantization not enough, then the camera start using the fallback matrix instead of the main matrix. this way it is possible to have two different matrices working together, the camera automaticaly changes between them.

@Manicd in theory yes, but in real world it cannot work so good for iso 160-2000 because the camera can activate the fallback when you do not want it, for example in a small motion inside the image which does not need a higher quantization, or in iso 800 + little motion, or in 1600 iso in low light, so It will not be so good as the version without the fallback. the version without the fallback for 24p can make the camera always use the lowest quantization matrix to get always the best noise texture in iso 2000 and below. in few words: the camera does ther change from main matrix to fallback matrix automaticaly, and automatic things are not so good all the time. the 24p version without fallback, for iso 2000 and below, is just amazing, looks like intraframe, great for the big screen, and minimizes the green cast.

apefos

@Manicd you can see that in the version for iso 2000 and below, only the 24p does not have fallback. this is because the datarate in 24p allows it to use same matrix up to iso 2000. the hbr uses the same 24p quantization for main matrix, but uses a fallback to handle the noise because the datarate does not allow the small quantization all the time. the 720p cannot use same small quantization for the main matrix due to more frames and small datarate, so it is another combination. I tested all them to do the best main+fallback combos.

matt_gh2

@apefos @producer Keep up the great work guys. Image looks very interesting. I will test "The Patch" when I get a chance. @producer Are you going to release your Filmvision Pro Plugins soon?

Manicd

But now I look at the text settings for seti iso2000 and setj iso8000. There is no difference in the files except for fallback of 1080i/30p and 720p. For 24p, there is no difference in settings between seti iso2000 and setj iso8000? Same with seta and b.

apefos

@Manicd

i and j have higher datarate compared to a and b, for people who wants to be in the bleding edge for datarate. please see txt file inside the zip and the yellow box in ptool, or just read the ini file.

the zip called "...4versions" does not have fallback for 24p in iso 8000, so I did another upload...

there is a post above when I say "iso 8000 and below versions, same as previous upload, but with fallback for the 24p:" this post has the 8000 iso, same versions with fallback for 24p, (I uploaded B and J again with fallback for 24p).

Manicd

Yes I compared all txt files. I apologize I'm still learning how the matrix/hack settings/and everything else works together.

What I see is:

Before 24p fallback, the 24p had same txt file settings for iso2000 and is8000.

Now with 24p fallback, setj iso8000 uses the setg iso2000 matrix as fallback.

apefos

yes, that's it.

apefos

to avoid mistakes and confusion, the correct final versions are here:

producer

The same test from above re-edited with avoiding the grain added accidentally by FilmVision Pro from using it with previous settings. Now the ISO6400/8000 is what it is. Some grading changes too. (Watch in HD):

kevin_kirchman

Wow, great job @apefos ! I am loving this patch. Here is the mandatory cat picture. Taken from a 24H video clip, graded in Speedgrade.

apefos

thanks for the test, I am happy you like it.

apefos

I like the 6 to be the dct number. To my eyes it is a good balance for noise/texture, sweet spot between fine/gross. More fine than FlowMotion which uses 7 as dct number, more texture than Nebula which uses 4 as dct number.

Some people likes Nebula due to it's more fine texture, also some people likes Smooth Cinema. Driftwood and cbrandin, both likes 4 as dct number, so I created new matrices based on my previous desing, but using 4 as dct number:

(as you can see, these matrix corresponds to the 4 matrices used in the final versions of The Patch, but using 4 as dtc number and proper frequency increase, similar to my previous matrices with 6 as dct number. I think these matrices will allow to create the 1600S, 6400S, 1600H and 6400H Patches, need to see which isos the matrices and fallback will handle ok.

Matrix 1:

004 006 008 010

006 010 012 014

008 012 016 018

010 014 018 022

Matrix 2

004 008 012 016

008 014 020 022

012 020 024 028

016 022 028 034

Matrix 3

004 008 012 016

008 014 020 024

012 020 028 032

016 024 032 048

Matrix 4

004 008 012 016

008 014 020 048

012 020 056 064

016 048 064 096

tcarretti

OK, I've been away from the scene for quite a while and am very intrigued by this thread. The last patch that I had been running was the Sanity 5.1 patch which was very stable and extremely usable. I'm curious to know if this patch would be reasonably stable using a Sandisk extreme 30MB/s Class 10 card. I don't know if I have any faster cards around. Any advice would be greatly appreciated. Congratulations regardless as it seems that you have created quite a popular patch!

Tony

Howdy, Stranger!

Categories

Tags in Topic

Top Posters