Regarding my restoration of Cutie Honey (1973)

[Troc]

Hello, first post and already pretty complicated. I however did search around and found fragments, not a specific discussion on this.
I'm a huge fan of Cutie Honey (1973), so I bought the DVD boxset. Being from the 70s, it has many issues which I want to fix with Virtualdub, since that's what I'm comfortable with. I think it was shot into film from the original source, which introduces all sorts of problems. So far, I'm using methods that might be suboptimal and some I cannot figure out correctly. Here we go:

- There's noise, which is a result of imperfect capture.
I use Neatvideo, which makes some really impressive results. It however does struggle a bit with digital artifacts and cannot remove some deeper film scratches/hairs.
- There's slightly warped duplicate frames, which again is a result of the film camera.
I tried MPDecimate with FFMPEG, but it didn't work for most duplicates, since in most of them, a small element or area sort of bubbles out, making the frame not identical, even though I can tell that it is the same frame. Removing the slightly altered duplicate frames manually would be horrendously slow. Do you have a solution?
- The frame shakes, which is a result of the film camera and the animation not being in perfect sync.
I tried using Deshaker, but I ran into problems where there either was extra black area (no inpainting?) or the frame was zoomed in with extra digital movement. I would like to have it work, since this is somewhat annoying.
The colors are washed out and yellowed.
I have altered saturation, value and hue and also tried to alter the white balance. However, it appears that some shots have different values and a static change makes them look weird when the rest of the shots look good. Do I need to separate those shots and alter them individually and later add together again?
- There's some background noise in the audio.
I would like to clean the audio up a bit. I cannot use the lossless audio from original files in Audacity since it doesn't have a library or something and I couldn't get it to work. If Virtualdub can do audio, what filter/settings should I use?

I could post a Mega or Gofile link to unaltered parts of the show, if it helps with your assessments.

So far, I can make the show much more visually pleasing. I know that I could make it even better, but my knowledge is limited and I learn best with interaction.

[StainlessS]

- There's slightly warped duplicate frames, which again is a result of the film camera.
I tried MPDecimate with FFMPEG, but it didn't work for most duplicates, since in most of them, a small element or area sort of bubbles out, making the frame not identical, even though I can tell that it is the same frame. Removing the slightly altered duplicate frames manually would be horrendously slow.
Do you have a solution?

Avisynth script function Spotless() might be a solution to above.
EDIT: Not for anime, dont know what your clip is.

EDIT: https://forum.doom9.org/showthread.php?t=181777

EDIT: Looks like "Cutie Honey" is manga, so not of use, sorry. [EDIT: although you might try with a very large BlkSz ]

[Troc]

It's actually several different things, but I'm talking about the anime series from 1973. I should probably give Avisynth a go, taken how many people seem to love it.

[FranceBB]

Yes, come to the dark side.
Ok, jokes aside, uploading a small sample would be helpful for us to try to see and get rid of those issues. Especially if you're not familiar with Avisynth, we could send you a script to take care of those things so that you just run it and then forget about it. You can even open Avisynth Scripts in VirtualDub (since that's what you're familiar with) and encode from there, it's totally fine.

p.s for what it's worth, I've been using SpotLess() in some of my scripts and it works, however all of them were real life sources, so I have no idea how good/badly is gonna behave with anime. I guess we're about to find out.

[Troc]

Originally Posted by FranceBB

Yes, come to the dark side.
Ok, jokes aside, uploading a small sample would be helpful for us to try to see and get rid of those issues. Especially if you're not familiar with Avisynth, we could send you a script to take care of those things so that you just run it and then forget about it. You can even open Avisynth Scripts in VirtualDub (since that's what you're familiar with) and encode from there, it's totally fine.

p.s for what it's worth, I've been using SpotLess() in some of my scripts and it works, however all of them were real life sources, so I have no idea how good/badly is gonna behave with anime. I guess we're about to find out.

Hello, thanks for responding. I agree that video restoration seems to be quite the dive.

Here's a sample, the OP: https://mega.nz/file/iCZjHIpa#Ogt0Wm...mDzffZlLMRyKFo

How would I actually run Avisynth scripts in Virtualdub? I can't seem to import and .AVS as a filter. I would really like to know, since there's all sorts of intesting Avisynth scripts/script combos that people tell me to use. I'm too dumb to get anything but "Version()" to run, though.

Hopefully you can use the video I sent.

[videoh]

Originally Posted by Troc

How would I actually run Avisynth scripts in Virtualdub?

Simply use 'File/Open video file' as you would to open an AVI file. Then just select your AVS script as the input file and off you go. The video generated by the script will appear. Off course, Avisynth must be properly installed.

[Troc]

Do I need to include the video path in the Avisynth script or can I apply a script to any video that's open already?

[videoh]

Your script will include a source filter that specifies the input file.

[Troc]

How would I make a script work? I get issues for not having the scripts stored in the right places, I think.

[videoh]

I don't understand your question. Please tell us exactly what you did and the exact problem you are encountering, together with your script contents and with error messages.

The process is simple:

1. Install Avisynth.
2. Make a script.
3. Open it in VirtualDub2.

[Troc]

The problem I encounter is "module not found" on every script element.

[videoh]

You did not provide the information asked for, so I cannot help you.

[Emulgator]

Troc, your sample is in no way the original DVD encode.
The artifacts are already dug in too deep.
Besides, it is .hevc. Better upload a cut of the original .vob

[Troc]

@videoh, here's the error I get. I put the avisynth script into Virtualdub and got this: https://i.ibb.co/Z27YdfZ/avisynth-error.png
I remembered it wrong, sorry.

@ Emulgator, I'm sorry, I don't know how to cut VOBs without using a codec. I like HEVC as a codec. Now I copied one VOB directly from the DVD and uploaded into my Mega. Here's a link. Hopefully it helps you:
https://mega.nz/file/Ta4kzBIR#nPyMpK...DQulFw0w7h89Ic
Fair warning though, the quality is poor.

[videoh]

Troc, you have to give us your script!

[Troc]

My bad again, it's here:

"C:\Users\Troc\Music\Be My Baby - Koda Kumi - Live.mp4"
nnedi3_rpow2(4, cshift="Spline36Resize", fwidth=2556, fheight=1440)
aWarpSharp2(depth=10)
Sharpen(0.5)

Any help? Removing nnedi resulted in the same error with aWarpSharp instead.

[StainlessS]

@ Emulgator, I'm sorry, I don't know how to cut VOBs without using a codec

In DGIndex, mark start and end of about 30 second sequence with lots of movement, using
'[' and ']' marker buttons, then
Menu/File/"Save Project & Demux Video",
and post the m2v file ONLY. [SendSpace allows post without account, file deleted 30 days after latest download].

[videoh]

Originally Posted by Troc

My bad again, it's here:

"C:\Users\Troc\Music\Be My Baby - Koda Kumi - Live.mp4"
nnedi3_rpow2(4, cshift="Spline36Resize", fwidth=2556, fheight=1440)
aWarpSharp2(depth=10)
Sharpen(0.5)

Any help? Removing nnedi resulted in the same error with aWarpSharp instead.

1. Your first line is wrong. You don't just state a path like that. You have to use a source filter.

2. You need to load all the (non-internal) filters that you use in your script using LoadPlugin(). There is also autoloading as a possibility. But one way or another you have to load your filters.

[Troc]

Originally Posted by videoh

1. Your first line is wrong. You don't just state a path like that. You have to use a source filter.

Alrighty, I think it's very clunky that I have to specify both the source path and use some different script just to use a file type. I think it's obvious that if I select a .mp4, I want to use what works best for .mp4. Why would I do something different?

2. You need to load all the (non-internal) filters that you use in your script using LoadPlugin(). There is also autoloading as a possibility. But one way or another you have to load your filters.

Why is autoloading not the default? I have a good SSD. The filter filesizes are nothing, I could load a thousand filters in a second. If that was not the case, where am I supposed to put the filter files I downloaded and how am I supposed to load them? It's again really clunky to not load a filter based on how I tell the program that I want to use a specific filter.

When I post stuff about video restoration, people like to tell me that they use Avisynth and how it's better than anything I know. A how to use but when I ask how, they link the wiki. As if I haven't tried to read it without getting anywhere. I want a tool, not an encyclopedia. Is there not a handy GUI for Avisynth with every filter already integrated? If that's possible, why's a newcomer like me either directed to a wiki or given some handwavy "just specify a source filter". If I knew how to do that, I would've done it and wouldn't be asking why it doesn't work.

So, I'm asking thoroughly. Let's say that I choose an Ffmpeg source filter.
http://avisynth.nl/index.php/FFmpegSource

I have to download the DLL and put it somewhere. Then, to start my script, I'll write the following: File path, if I want it cached, write the fps amount, how many threads the CPU encoding will use, the length of the video, how to seek the video, if I want to overwrite index, the video's resolution, resizer if necessary, if it's RGB or YUV, RFF (whatever that is), delay, and set UTF8 if I have a strange filename.

After I do that, I have inputted a singular file to then apply filters to. Then I select a fiter, like http://avisynth.nl/index.php/TIsophote
I separately specify all the parameters with the filter and try to export multiple times to see what actually looks good. Or do I have to use an output/export filter as well to make this thing work as it should?

I'm sorry for the extreme levels of salt I have. I'd like to think that I'm a relatively intelligent person with some skills in video editing. Avisynth however is something I do not understand. It's maddening since so many people seem to use it effectively.

Originally Posted by StainlessS

In DGIndex, mark start and end of about 30 second sequence with lots of movement, using
'[' and ']' marker buttons, then
Menu/File/"Save Project & Demux Video",
and post the m2v file ONLY. [SendSpace allows post without account, file deleted 30 days after latest download].

I had never used DGIndex. It seems quite effective, however. Here's the file: https://www.sendspace.com/file/vo9apj
I used MakeMKV to rip the DVD before.

[videoh]

Originally Posted by Troc

Alrighty, I think it's very clunky that I have to specify both the source path and use some different script just to use a file type. I think it's obvious that if I select a .mp4, I want to use what works best for .mp4. Why would I do something different?

Sorry to say it but that is nonsense. Here is a way to correctly write your first line using (for example) DGDecNV tools:

DGSource(""C:\Users\Troc\Music\Be My Baby - Koda Kumi - Live.dgi")
# now all your other filters go here

You create the dgi file (called the index file) by indexing the MP4 file in DGIndexNV. You can use other source filters, some of which will index on-the-fly, and some of which do not index at all, or require manual indexing as with DGIndexNV. Indexing allows for random seeking. There is also a version of the DG tools that does not require an nVidia GPU, that is, DGIndex/DGDecode. They both work the same.

So the basic pattern is simple:

YourSourceFilter("path_to_video_or_index_file", ...other params as needed...)

Why is autoloading not the default?

Nobody said it isn't. But you need to have Avisynth installed correctly and place your plugins in the right place.

where am I supposed to put the filter files I downloaded and how am I supposed to load them?

If you don't want to use autoloading (and there are good reasons for sometimes not wanting to), then you put them wherever you want and then load them in your script with LoadPlugin().

I want a tool, not an encyclopedia.

Gee, stop whining. All tools come with user manuals. Avisynth's user manual is a wiki.

why's a newcomer like me either directed to a wiki or given some handwavy "just specify a source filter". If I knew how to do that, I would've done it and wouldn't be asking why it doesn't work.

You've got to be kidding. The wiki contains many examples that show how to use source filters.

Or do I have to use an output/export filter as well to make this thing work as it should?

No, you simply open your script in the desired application. As I mentioned earlier you can open the script in VirtualDub2 if you just want to see the filtered video.

I'm sorry for the extreme levels of salt I have. I'd like to think that I'm a relatively intelligent person with some skills in video editing. Avisynth however is something I do not understand. It's maddening since so many people seem to use it effectively.

You're overthinking things. I gave you the basic 3 steps earlier. Make a script with just a source filter invocation and then open that script in VirtualDub2 (File/Open video file). You'll see the source video. Now add another filter like Sharpen(). Now when you open the script you'll see that same video but sharpened.

I had never used DGIndex. It seems quite effective, however.

Thank you, I appreciate that.