(Noob) Deinterlacing issue with early 2000s DVD-Rip | AviSynth + QTGMC

[expo_]

Hi everyone,

Im very new to AviSynth and QTGMC, and I have been trying to deinterlace some old childhood DVD Rips (Hot Wheels World Race and AcceleRacers Animated Movies from 2003-2006).

Originally I used handbrake but after hearing about QTGMC I decided to give it a try in hopes of achieving less picture damage and just better results in general. I found and followed this tutorial https://youtu.be/C4PyyQoz6eo?si=Apt-6JNHuR3g0ECL.

My process has been ISO > MakeMKV > run script on MKV file using AvsPMod

Script

SetFilterMTMode ("QTGMC", 2)
FFMPEGSource2("AcceleRacers 2.5 [Ice Realm].mkv", atrack=1)
AssumeTFF()
QTGMC(preset="Slower", EdiThreads=3)
Prefetch(10)

when viewing the preview I see repeating frames that jump back and forth throughout the video. I'm not sure what could be causing this, for reference, here is the mediainfo on the MKV file I am working with.

[Attachment 75560 - Click to enlarge]

Here is the problematic video https://drive.google.com/file/d/1k3L2Fa7SJiV5ZwVNpYlwikRbzCmL4Tkt/view?usp=sharing.

and here is the original mkv from makemkv https://drive.google.com/file/d/1pr9cJYBrNkHOeACvroRoK9A1kDPJzZim/view?usp=sharing.

EDIT: These are my export settings within virtualdub2

[Attachment 75573 - Click to enlarge]

[Attachment 75574 - Click to enlarge]


Something is going wrong and I cant figure out what it might be, I'll be happy to provide any additional details that may help.
Any help would be super appreciated, thank you!!

[Alwyn]

If you have jumping backandforth frames, it could be that your DVD is bottom field first. Try BFF instead of TFF.

Otherwise, please post the original MKV here as an attachment (max 500mb) or host it on a file sharing site such as Google Drive, Dropbox. YT re-encodes it's video so it becomes much more difficult to diagnose/suggest fixes with backed-in errors.

[expo_]

Just updated the post to include a google drive link to the original MakeMKV Rip. If there is anything else I can add lmk.

Also, I gave BFF a try.. no luck. When using TFF, everything after the intro was alright, but now with BFF the entire thing is jumping back and forth same as the intro.

EDIT: After closer examination, the jumping is present throughout the entire video, not just in the intro.

[Alwyn]

That's great, the experts can now do their stuff.

[Sharc]

It is telecined and should therefore be Inverse-Telecined:

LWLibavVideoSource("AcceleRacers 2.5 [Ice Realm].mkv")
assumeTFF()
tfm().tdecimate()

[lollo]

It is not an interlaced source

edit: Sharc arrived first!

[Alwyn]

It is not an interlaced source

There's what looks like heavy interlacing at frame 178, which is largely cleared by Double Rate deinterlacing. I also see Sharc "assumes TFF". Does that not indicate the file is interlaced? Could you explain?

[Sharc]

Telecine looks like interlaced as it uses interlacing technology, but it is not the same as true interlacing. You see a repetitive sequence of N clean frames followed by M combed frames.
https://en.wikipedia.org/wiki/Three-two_pull_down

[Alwyn]

Thanks Sharc, is your Assume TFF necessary then?

[lollo]

There's what looks like heavy interlacing at frame 178, which is largely cleared by Double Rate deinterlacing. I also see Sharc "assumes TFF". Does that not indicate the file is interlaced? Could you explain?

Because you used a deinterlacer on the source to analyze it, which is technically wrong, because the deinterlacer makes its own conclusions and processing.
Is better to analyze the original untouched fields to understand the complete architecure.

# plugins directory
plugins_dir="C:\Users\giuse\Documents\VideoSoft\MPEG\AviSynth\extFilters\"

	# FFmpegSource
loadPlugin(plugins_dir + "ffms2_87bae19\x86\ffms2.dll")

video_org=FFmpegSource2("AcceleRacers 2.5 [Ice Realm].mkv", atrack=-1)

# separate fields tff
video_org_sep_tff=video_org.showFrameNumber().AssumeTFF().separateFields()

	# to check if progressive or interlaced
return(video_org_sep_tff)

By opening this script in VirtualDub and moving frame-by-frame you see that the even and odd fields of the frame you mentioned refer to different moment in time, like in interlaced fields, leading to a "combing" in that frame; but there is no overall "interlacing" across the whole structure.

[Sharc]

Thanks Sharc, is your Assume TFF necessary then?

It's not needed for clean telecine patterns without breaks in the sequence. It may help for poorly edited telecined stuff with breaks in the pattern and orphaned fields.

[Alwyn]

Thanks Lollo, Sharc. I'm always learning!

[expo_]

Tysm! One final question about it all, there are pal releases of the films.

Would it be better to get ahold of those PAL discs which are in original 25 fps progressive and I assume had been telecined to meet ntsc's 29.97fps?

[Sharc]

Thanks Lollo, Sharc. I'm always learning!

Btw. you can Inverse Telecine in Vdub2 using its IVTC filter:

[Sharc]

Tysm! One final question about it all, there are pal releases of the films.

Would it be better to get ahold of those PAL discs which are in original 25 fps progressive and I assume had been telecined to meet ntsc's 29.97fps?

Yes. It's usually always better to work with the originals.

[Alwyn]

you can Inverse Telecine in Vdub2 using its IVTC filter:

I did try that (using different settings to you) but even with your settings, the AVISynth result is miles better.

My VDub attempt:

[Attachment 75593 - Click to enlarge]

Your AVISynth (ignoring aspect ratio):

[Attachment 75594 - Click to enlarge]

[Sharc]

you can Inverse Telecine in Vdub2 using its IVTC filter:

I did try that (using different settings to you) but even with your settings, the AVISynth result is miles better.

Avisynth's TIVTC deinterlaces orphaned fields (those for which no match is found) by default. Study its options for TFM() and TDecimate() for best results. Usually the defaults are fine though.
http://avisynth.nl/index.php/TIVTC/TFM

The patterns of Anime are often more tricky to analyze than "natural" footage.

[expo_]

Thank you all for helping, I really appreciate it.

Is there any way to export audio along with the video using LWLibavVideoSource? With QTGMC I was able to put ",atrack=1" after the input files name. I can add the track in seperately after everything using MKVToolNix, but its another step.

[Sharc]

Thank you all for helping, I really appreciate it.

Is there any way to export audio along with the video using LWLibavVideoSource? With QTGMC I was able to put ",atrack=1" after the input files name. I can add the track in seperately after everything using MKVToolNix, but its another step.

Try something like

A=LWLibavAudioSource(....)
V=LWLibavVideoSource(....)
AudioDub(V,A)

[hello_hello]

If you output the audio with Avisynth it has to be re-encoded, whereas MKVToolNix doesn't re-encode.

There's two functions here that do much the same thing as FFMPEGSource2 (combining the video and audio from FFVideoSource and FFAudioSource). Most versions of FFMS2 can also use FFMS2("source file.mkv") as a function instead of FFMPEGSource2.

One of these functions is for LSMASHVideoSource / LSMASHAudioSource and the other is for LWLibavVideoSource / LWLibavAudioSource.
http://avisynth.nl/index.php/LSMASHSource#Examples

Copy the text and paste it into a text editor, then save each one with an avsi extension and put them in the Avisynth plugins folder. The function names are LSmashSource2 and LibavSource2.