Anyway to restore all frames from NTSC>PAL animation

[nick1977]

Hi all

Wherever possible (assuming there is no Blu-Ray available) I will buy NTSC sourced DVDs for animated series as more often than not they can be detelecined to provide a decent 480p video. However sometimes there is no option other than to buy them from UK which means they are 576i50. Now inlike films, animated series don't tend to have PAL speedup, instead relying on a crude frame conversion. Deinterlacing these clips even to 50fps still leaves some juddery ghost frames.

I'm guessing it's impossible to get back to the native 23.976fps but what is the best way to get closest to this? I've played around with virtualdub quite a bit but never made the jump to avisynth.

Interested to read people's thoughts!


Nick

[jagabo]

Can't be done in Virtualdub. In AviSynth:

WhateverSource()
QTGMC() # or Yadif(mode=1)
SRestore()

[nick1977]

Can't be done in Virtualdub. In AviSynth:

WhateverSource()
QTGMC() # or Yadif(mode=1)
SRestore()

Thanks for the quick response
Happy to invest some time in avisynth if it's worthwhile.
Does this script provide perfect results, or is it just a question of being "less bad"?
Are there any sample clips out there anywhere?

Thanks

[poisondeathray]

Does this script provide perfect results, or is it just a question of being "less bad"?
Are there any sample clips out there anywhere?

If your description was correct (there are other types of conversions), and the source used typical field blending (not edited afterwards), it produces close to perfect results in terms of cadence and original frames. So not just "less bad"

There are many examples out there posted at doom9 and on this forum. If you want to post an unprocessed sample clip, somebody will run the same quickly and you can see for yourself

[jagabo]

Happy to invest some time in avisynth if it's worthwhile.

QTGMC is one of the hardest filters to get started with because you have to download and install several other third party filters. The QTGMC page lists them.

http://avisynth.nl/index.php/QTGMC

There are also a links there to archives that contain all the needed filters. See the Plugin Packages section.

Does this script provide perfect results, or is it just a question of being "less bad"?

The results are usually very good. A source that has been through a single NTSC/PAL conversion with field blending will get nearly perfect results.

Are there any sample clips out there anywhere?

search through the forums here for SRestore. You'll find lots of examples.

[nick1977]

Great thanks guys.
I have a 30 second clip I can post which features some panning scenes. What format should I upload? The original is a rip from a home recorded dvd so 25i 704x576 mpeg2.

I'll try and upload later this evening when I get home

Nick

[jagabo]

Upload the original MPEG 2 video. You can use a program like Mpg2Cut2 or DgIndex to extract a sample.

[nick1977]

I've attached a short clip for testing. Hopefully someone will be kind enough to have a play around with it.
As stated previously, my target for this is to achieve 23.976p. Clips from the same show that I have sourced from NTSC DVD de-telecine perfectly.
Let me know if a different / longer clip is required

Thanks in advance for any help!

Nick

[jagabo]

You may want to do some more cleanup but here's the result of just QTGMC() and SRestore() to remove the blending and restore the original 23.976 fps frame rate.

Mpeg2Source("VHStestclip.d2v", CPU2="ooooxx", Info=3) 
QTGMC(preset="fast")
SRestore()

[nick1977]

Many thanks for testing for me.

I've watched it back and certainly the first section of panning movement looks almost flawless - far better than I've been able to achieve with virtualdub. However the second pan (the last few seconds of the clip) seems to stutter quite a bit. Is this due to it being such a short clip, or is this inconsistency a common result?
If I could get results consistently like the first half of the clip, I can see it's definitely worth me investing the time to get it working for myself

Thanks again

Nick

[jagabo]

Sorry, I saw that it was working with the first part of the video and didn't even watch to the end. The technique here isn't perfect but keep in mind that a lot of cartoons are shot on film, telecined to video, then edited. Shots are often slowed or sped up while editing -- using the same field blending techniques that are used for NTSC/PAL conversions. When that is done there's no way to restore the entire video to a single frame rate. That may be the case here. I would try it on a full length episode and see what proportion has the problem.

By the way, you can change the frame rate output by SRestore by including frate=X as an argument. For example:

SRestore(frate=23.7)

Getting the right value (23.7 is close) will smooth out that portion of the video, but others will get jerky.

[nick1977]

Great, thanks very much

[nick1977]

Sorry I've been thinking about this some more and I'm wondering if the source animation is actually the issue. I accept that there is a possibility something could differ in the sources, but it does seem odd that all of the NTSC sourced episodes work perfectly with an inverse telecine in handbrake, whereas this PAL sourced clip potentially has variable framerate animation. The animation rate wouldn't be affected by the conversion from NTSC to PAL would it? It is an episode midway through a season and episodes either side (sourced from NTSC DVD) are all fine after inverse telecine in Handbrake.

Hope that makes sense!

To put into context, I have roughly 100 episodes from NTSC and all of them appear fine after an inverse telecine.

[jagabo]

Sorry I've been thinking about this some more and I'm wondering if the source animation is actually the issue. I accept that there is a possibility something could differ in the sources, but it does seem odd that all of the NTSC sourced episodes work perfectly with an inverse telecine in handbrake, whereas this PAL sourced clip potentially has variable framerate animation. The animation rate wouldn't be affected by the conversion from NTSC to PAL would it? It is an episode midway through a season and episodes either side (sourced from NTSC DVD) are all fine after inverse telecine in Handbrake.

Hope that makes sense!

To put into context, I have roughly 100 episodes from NTSC and all of them appear fine after an inverse telecine.

This video was obviously made from a VHS tape. There's no telling how what kind of conversions it's been though and how badly it's been mangled.

The realtime hardware PAL/NTSC converters (often used for analog NTSC to analog PAL conversion) do not change the running time. They change the frame rate by adding or removing fields and blending them when field durations overlap.

[nick1977]

OK got you now
This will no doubt seem a bit cheeky so fully understand if you're not happy to, but if I attach a full episode (approx 5 min duration) could you run the script on that? I could then see if it's the odd issue or if it will be "off sync" a lot. I am not able to spend a lot of time doing scene by scene conversions so need to see if the result overall is worth learning the avisynth process for or whether I'll just stick with my current processing method.

Many thanks

[nick1977]

OK got you now
This will no doubt seem a bit cheeky so fully understand if you're not happy to, but if I attach a full episode (approx 5 min duration) could you run the script on that? I could then see if it's the odd issue or if it will be "off sync" a lot. I am not able to spend a lot of time doing scene by scene conversions so need to see if the result overall is worth learning the avisynth process for or whether I'll just stick with my current processing method.

Many thanks

[jagabo]

I'm willing to convert the whole episode for you. But AviSynth is well worth learning. There's all kinds of stuff you can do with it that can't be done (or at least, not easily) with other software.

[nick1977]

OK thank you.
Yes I am sure it is worth learning. I have read about it on numerous occasions, just always fought shy of it as it appears quite complex!
I've attached the full clip this time, so look forward to seeing how it does.

Thanks

[jagabo]

Here's the whole clip with just QTGMC().SRestore(). I think the jerky shot is a little better. There are still a fair amount of blending when Butthead (?) is rubbing the sticks together.

[nick1977]

I'm VERY impressed with that conversion. My sincere thanks for talking it through and providing the test runs.
I'll have to set some time aside to get my head around avisynth before I continue with this backing up "project"!
One quick question. I assume it will not affect the results of the conversion if the input is an uncompressed avi which I extracted from the DVD? I only ask as I have already split out a number of the episodes, and they are simply an uncompressed avi saved from virtualdub. No processing at all.
If there's a chance they will be affected I'll go back to the source DVD, but just thinking it would save me some time!

Thanks again

[jagabo]

VirtualDub can mess up interlaced YV12 video. If you did nothing but open the MPEG 2 source, mark-in, mark-out, and save as uncompressed YV12 AVI the video is probably intact. If you let VirtualDub convert to RGB or YUY2 it probably messed up the chroma channels. In that case you'd be better off going back to the DVD source.

[lordsmurf]

I'd have to see the source to know what's possible. Maybe I'll look at the thread attachments later.

I'm currently restoring 100+ episodes of a Japanese PAL-only-released cartoon with issues. After 20 years, I've given up on a proper release, and the source I now have is good enough to be mostly restored. It's all Avisynth, mostly SRestore (which switches, not default).

jagabo is spot-on with his advice here.

[nick1977]

Sorry to drag this up after a little time.
I've finally got around to trying this for myself, and keep hitting multiple stumbling blocks with virtualdub and avisynth.

What I do know is that I can create a script file which works with version() test.
I can also input from a d2v file that I created, using the mpeg2source.
The trouble comes when trying to use other functions named above QTGMC and SRestore
I have tried turning both on and off with a # as both error out with a msg referring to a line number in the function.

I've tried so many version of vdub and a couple of avisynth and I now feel like I'm going around in circles.

Can someone suggest version numbers etc of virtualdub and avisynth that I should be using please?
I am on Windows 10.

Thank you!

[sneaker]

QTGMC has a lot of requirements. SRestore() also has some. Install them as they are shown in the error messages. Once you have finished that try again. If there are still error messages after that post your complete script and post the exact error message.

AviSynth: use the standard 2.6.0 version (32 bit).
VirtualDub: use the standard 1.10.4 version (32bit).

[nick1977]

QTGMC has a lot of requirements....

I noticed...!

If there are still error messages after that post your complete script and post the exact error message.

OK, I have installed the versions linked to and I now have this issue:

Script error: there is no function named "mt_makediff"
(QTGMC.avsi, line 918)
(QTGMC.avsi, line 473)

I've googled and this seems to be linked to the masktools2.dll
I have the x86 version installed dated 27/04/2017 13:26

Any help appreciated!

[nick1977]

Sorry for forgot to post the code (It was very late!)

It's very basic really - just trying to initially do what jagabo did above:

MPEG2Source("e:\a\godz.d2v") 
QTGMC(preset="slower")
SRestore()

I've tried removing the preset option, and putting fast, but it makes no difference.

Thanks

[sneaker]

Try installing Microsoft Visual C++ 2015 Redistributable Update 3.

[jagabo]

mt_makediff() is part of mt_masktools 2.

[nick1977]

I've run the C++ installation. I had masktools2.dll in the plugins folder, but re copied it just in case.
Still no joy


masktools2.dll is version 2.2.8.0 and I used the x86 non XP version.

I'm sure this shouldn't be this hard!

[sneaker]

Did you restart the PC?

Run avsmeter -avsinfo -l and post the log file here.