![]() |
[HELP] VMAF comparison - Printable Version +- Selur's Little Message Board (https://forum.selur.net) +-- Forum: Hybrid - Support (https://forum.selur.net/forum-1.html) +--- Forum: Problems & Questions (https://forum.selur.net/forum-3.html) +--- Thread: [HELP] VMAF comparison (/thread-4057.html) |
VMAF comparison - ThePower - 01.04.2025 Hi I am new and I'm having trouble using Hybrid I have a DVD that I have ripped for backup using MakeMKV. I extracted the individual episodes from the ISO into mkv files. I wanted to make 60fps QTGMC versions so that it looked more like it did when it was on a CRT TV back in the day at 60 fields/per second. So far, so good. Step 1: Reencoded the mkvs into FFV1 with QTGMC+Bob (to make a 60fps video as losslessly as possible. I say losslessl(y) because QTGMC is running. I kept dupicates frames so i didnt have to adjust any audio/video sync issues, HDD space is plentiful.) Step 2: Reencoded the resulting files with SVT-AV1 using preset 2 CRF 9 to get the best Quality/Time/FileSize ratio so I can stream to another device. Im okay with it being double the original 800mb file considering its double the framerate, the FFV1 would be deleted at this point and I would consider this the new "original" qtgmc copy from DVD with a VMAF above 97%. I was wondering, am I doing this right? I wasn't sure if reencoding straight from the episode source mkv to AV1+qtgmc would allow me to skip the FFV1 encode...I guess Im confused as to "when" hybrid uses qtgmc in the encode process and whether or not it makes a difference.. Because of this I decided to do a test. I compared the VMAF of the (Source mkv)>(FFV1 lossless+QTGMC)>(SVT-AV1 preset 2 CRF 9) to (Source mkv)>(SVT-AV1 preset 2 CRF 9 +QTGMC). File sizes are 358,089,338 and 358,089,332 bytes respectively, with a VMAF of 92%. I can't figure out why there is a file size discrepancy of 6 bytes. The file sizes are close enough I figured maybe it had to do with tags, file names, etc, and that the videos themselves were 1:1 identical. I expected a 100% VMAF score when comparing them. No settings were changed during the AV1 encodes. The only difference is one AV1 came from a FFV1+QTGMC lossless encode of the 30fps dvd mkv file, the other came straight from the 30fps dvd mkv file using QTGMC. Shouldn't they be identical? What am I missing? Too Long : Didn't Read: 1. Im trying to get a VMAF of >=97%, 60pFPS using qtgmc+bob, with a file size of roughly double. I have settings set so that it looks "as close as possible" to how it looked on a CRT back in the day to my well trained eye, but on modern displays. 2. Do i need to encode to FFV1 to use qtgmc first, or can i just go from source>av1+qtgmc? 3. They clearly generate different outputs, so what would be better/how to get them to be 1:1? 4. I compared two videos using VMAF and they did not come out as expected, which should have been 100%. If anyone asks why I don't just leave them interlaced, its because interlaced looks horrible on progressive displays. The CRT's phosphor glow decay time is hard to beat and quite literally magic, which I guess is why QTGMC was created in the first place. Sorry for the long post, and thanks for the amazing program! I cant believe it works on Linux using wine ![]() RE: VMAF comparison - Selur - 01.04.2025 1. VMAF between identical files often does not return 100% 2. File size differences even when remuxing the same content to mkv can be expected. 3. Encoding and filtering with multithreading often is not deterministic. I see no need for the intermediate FFV1 file in your workflow. Cu Selur |