Video Enhance v2.6.4

oh great! I had not seen that we could click on the codec in blue to select others… haha for qtgmc, it’s good, I reproduced my video correctly with vegas without deinterlacing with the gpu, the framerate is well doubled at 50 fps with qtgmc :slight_smile:

1 Like

Da Vinci Resolve Studio neural engine deinterlacer now tops QTGMC since the chroma combing artifacts with 4:2:0 videos has been fixed.
You can keep untouched the original grain/noise/sharpness (if exporting lossless) which is impossible with QTGMC no matter what settings you use.
Then using it with proteus in VEAI will give you superb deinterlaced restored videos without the ghosting artifacts the Dione models unfortunately still show.
QTGMC remains superior only with poor quality videos because of its stabilizer/temporal smoothing of compression artefacts. But for DVD quality or + Resolve is now better than QTGMC.

1 Like

really? can I download a trial version?

if you have 300€ yes :frowning: (the free version doesn’t have the feature needed for doing that)

As far as I know there is no trial version of the studio version and all Studio only functions are deactivated in the free version.

All right… well I’ll stay with qtgmc then… Laughing out loud.

Properly deinterlacing is complex. The main QTGMC thread on doom9 is insane. The documentation is insane. The script is insane but worth studying. I kept it open in Notepad++ for 6 months to understand the basics… Here’s a proper download of QTGMC that I started with:
QTGMC download

You’ll need the 32bit version of Virtualdub2 or 32bit version of whatever you load your .avs script with plus the 32bit version of Avisynth 2.6. Once you get the hang of it, you can install Avisynth+, get all the 64bit dlls and use 64bit version of Virtualdub2 (it’s slightly faster, quality is the same). If no one else does so, one day I may gather all the 64bit DLLs and post the collection to doom9.

Why in 2021 soon 2022 still using 32 bits ?
Avisynth 2.6 is now obsolete and everyone should exclusively use Avisynth+ and use 32 bits only with some old plugin that have never been upgraded to 64 bits only if necessary.
All QTGMC requirements have 64 bits versions
It’s not that complicated to install manually every plugin needed
http://avisynth.nl/index.php/QTGMC#Requirements

With 32 bits it’s not rare you get “out of memory” errors because QTGMC is very demanding. This never happens with 64 bits.

For deinterlacing 1080i, QTGMC is a lot easier than standard definition:
Your avs file:
01-cr1-2016.avs
DGSource(“cr1-2016.dgi”, deinterlace=0)
QTGMC( Preset=“Fast”, TR0=1 )

Load with Virtualdub2, set audio to none, set compression to Lagarith, Save as:
01-cr1-2016-qtgmc-fast-TR0=1-50fps-lag.avi

Load the .avi into VEAI.

Change your filenames to something meaningful to you. In my case, this was a 1080i 50fps MPEG2 video that I wanted as a H264 1080p 50fps. QTGMC + VEAI allows you to get excellent progressive video from an interlaced source.

Noteworthy is that it was true highdef video, meaning it was a real 50 fields per second. No fpsdivisor or SelectEven() needed. You have to determine that in advance with Virtualdub2 or an editor like VideoRedo by stepping through the frames and observing if each field is unique (motion in each field).

Some 50fps or 59.97fps 1080i content is actually 30fps. You’ll see each field repeated once (or more if telecined). In that case you need to decide it you want high frame rate or not. Sports and concerts look best at 50fps or 59.97fps IMO. Movies do not. They should be 23.976fps when you’re done for proper cinematic look IMO. I won’t get into telecined content (movies) as that quickly become complicated to explain.

If you need to double the framerate, there’s FrameRateConvertor for Avisynth. You need to use the slowest option to get decent quality. Video without credits can be frame doubled with SVPFlow (the old free v4.2.0.142) is fast and same quality as FrameRateConvertor. You need a NVidia GPU for it to work. The Avisynth commands to invoke it for frame doubling make no sense:
super=SVSuper(“{gpu:1}”)
vectors=SVAnalyse(super, “{}”)
SVSmoothFps(super, vectors, “{}”, mt=1)

Right, no logic to that syntax whatsoever.

Note I used DGDecode for my MPEG2 source. I created the .dgi file with DGIndex. You need that for QTGMC to work properly. It needs perfect frame accurate decoding of the MPEG file as it seeks back and forth as it works. Only DGDedode and VirtualdubMPEG can to that AFAIK. For interlaced H264, there’s DGDecodeNV. It’s $15 and you need a NVidia GPU.

Folks here have several different workflows. Some folks bypass the intermediate .avi step and drag and drop their .avs into VEAI to directly process the video. I have yet to try that.

Very interesting! Yeah the temporal smoothing QTGMC does sure doesn’t help standard definition get detail with VEAI. For 1080i, I doubt there’s any visible difference compared to my method.

How about you put that 64bit DLL collection together then? :slight_smile:
It took a week for me track down every 64bit version (I’m slow) and I totally did not understand Avisynth+ for months. The documentation does not explain where to put the DLLs files or how the different version get invoked.

For those wanting the Cliff Notes about Avisynth+:
Put the 32bit DLLs in the “plugins” folder. Put the 64bit DLLs in the “plugins64” folder. For .avsi files, copy to both folders.

The 32bit versions get invoked when you load your .avs with a 32bit editor. The 64bit versions get invoked when you load an .avs into a 64bit editor. NO WHERE ANYWHERE (but here now) is that explained to dummies like me.

For interlaced telecined film QTGMC is not recommended, you should use TIVTC (tfm + TDecimate filters)

i have a lot of concerts in bluray which are filmed at 1080i, that i would love to convert in 1080p. Your settings are not easy to follow, i never used avisynth, don’t know if i can do all that with staxRip, i’ll try to follow your settings. thanks for all these informations !

Preset selector is stuck on “Custom” after I make a change to my settings. Selecting one of the existing presets doesn’t change the drop-down selector to that preset, it remains at “Custom.”

This is not true. TR2=0 (final temporal smoothing = none), SourceMatch=3 (source match filter = 2x refined).

QTGMC(Preset="Placebo", InputType=0, TR2=0, Sharpness=0.2, SourceMatch=3, Lossless=0, MatchPreset="Placebo", MatchPreset2="Placebo", ediThreads=2)

(I don’t use “Lossless” because it isn’t actually “lossless”, it’s just called that, and it introduces artifacts.)

The above line is what I always use. I’ve read countless examples from people claiming to have better settings and I’ve tried them all, actually hopeful that I could get improved results, but the above line I find to be superior in all cases I’ve tried so far.

As you say “lossless” in QTGMC isn’t exactly true lossless and source match filter is just an additional filter process trying to match the original sharpness you can still manually tweak but still remains a process causing loss of some original details.
TR2 is for the final stabilization but TR0 TR1 Rep0 Rep1 Rep2 are still there and if you disable all of them you will likely get terrible results and you better stick with Yadif or use nnedi3 alone.

If you don’t want any QTGMC feature (stabilization, temporal smoothing, denoising) you should try nnedi3 alone with its defaults settings and compare with your QTGMC settings.
just change the field setting according to your desired framerate output and input carecteristics

     -2 = double rate (alternates each frame), uses avisynth's internal parity value to start
     -1 = same rate, uses avisynth's internal parity value
      0 = same rate, keep bottom field
      1 = same rate, keep top field
      2 = double rate (alternates each frame), starts with bottom
      3 = double rate (alternates each frame), starts with top

nnedi3(field=X)

Do you know how it compares to Hybrid? What do you use to create such lines?

Hybrid

I’m having a real problem with Proteus v3 producing unwanted moire patterns compared to v2.

2 Likes

What is the minimum size of the input video file?
21x25 does not work.

ps: I tried to find info on the forum, but didn’t.
ps2: sry for Google Translate. (●’◡’●) HNY