VirtualDub2

[nji]

Some time ago I discovered the Capture feature of VD2.
This came pretty handy as I was short before searching for
some other programs for that - for legal content of course.

It works pretty fine for me;
but with the time the little "stuttering" came to my attending.
Having I closer look there showed dupes and jumps in the capture.
For sure I have to do a matching between source and capture fps
(which I did by trial and error),
but even if it matches very well, there is say every 10th frame
a dupe followed by a drop. Especially when frame contents changes a lot.

I searched for information about how to handle that,
but nothing is specific enough.
Maybe somebody is able to answer my questions?

How do I know the frame rate of the source (some source stations
seem to change from strip to strip)?
Is there still a difference between F5 and F6?
(How) can the "Real-time profiler" be a help for that?
(I don't know the meaning of the statistics).
In this situation, is it better to use (low-CPU, large data) lararith codec
or say h264 (higher-CPU, less data)?
How can I know a potential bottleneck (memory etc.)?
My hardware is from 2018: i7-6700 (on-chip GPU, no graphic card etc.),
SSD for capture.

I would really appreciate if somebody has some help for me.

[isidroco]

Frame Size and FPS? Usually Lagarith using YV12 instead of RGB is fast and not so heavy, but HDD may be a bottleneck. Your CPU is quite fast, h264 with a fast preset may be better in keeping MB/s low. You should probably try using a standard FPS (regardless of dupes) of 25 or 30 fps.
Win10 may have a lot of background tasks/ services/ updates that may interrupt capture. You may try my automatic script/ utilities for that (OptimWin Batch on sourceforge).
What are you trying to capture? Maybe player is stuttering and not VDub, VideoLan VLC player might be faster than MPC. Open taskmanager to monitor what's taking more resources.

[nji]

Thank you for your hints. I will try them... one by one. (As you asked: The frame size I mostly use is 1280x1024 @ 24 fps, playback in browser). Another idea that came to my mind: if the source site offers it I could try lowering the playback fps, and record with that... to change it afterwards. I wouldn't have thought it may be as tricky nowadays...

[nji]

Recently I made a very strange oberservation:
Load a movie with VD2,
take filter MSU Denoising https://compression.ru/video/denoisi.../index_en.html
choose "Add filter" to open its dialog
(the CPU consumption instantly goes to 100% of 1 core)
close the dialog with cancel.
The CPU stays at 100% as long as you don't close VD2.

What is going here?
Bug in VD2, or is it From Russia with Love?

[LigH]

You suspect someone forging Bitcoins on your CPU time?

I don't know how much ProcessExplorer or similar SysInternals tools can help you digging deeper.

[nji]

I'm just interested about this quite unexpected effect.

[blob2500]

Originally Posted by nji

Recently I noticed that AAC in VD2 offers only CBR mode, while most prof movies use VBR mode.
So my (second) question:
Should I use VBR, and how to do that?
(I suspect I need an external codec for it?)

VD(2) doesn't support AAC natively to mux in AVI.
Support is via plugin:

Example: import through mkv(and mka) or mov plugin by fcchandler.

In this way, muxing mode is VBR always, even in case of constant bit rate aac audio into mkv/mka/m4a.

It' is similar (in avi headers level) to native mp3 vbr muxing import audio for mp3 (vbr or cbr). This works fine for MP3.

VBR muxing mode vs. CBR muxing mode is in AVI header, not in stream.

[LigH]

I don't see nji talking about saving AAC audio in an AVI container. Just above your quote he mentioned the MP4 container; this is where AAC audio belongs.

The integrated ffmpeg AAC encoder offers only an ABR like bitrate control; fdkaac does the same.
Nero AAC and QAAC offer real quality based VBR; VirtualDub2 may not be able to multiplex those without configuring an external MP4 muxer like MP4Box as "external codec", though.

[blob2500]

I didn't understand nji post, sorry

[nji]

Recently I noticed that a filter I use obviously reads/ writes the registry.
I thought about it...
Yes, filters obviously are allowed to read/ write files too (e.g. DeShaker).
So the next thought:
As they are not "sandboxed" like JavaScript or browser add-ons:
(How) can I trust a filter not doing malicious things?
(Especially if it is not open source).

[StainlessS]

(Especially if it is not open source).

Not sure if the filter you reference is Deshaker or not, but if so then is now open source.
Deshaker @ Github:- https://github.com/gu-t/Deshaker

[nji]

Yes, I know that.
It was just an example of what filters may do.

[shekh]

Originally Posted by nji

(How) can I trust a filter not doing malicious things?
(Especially if it is not open source).

You can`t completely trust it. Certain sensitive OS parts are somewhat protected from accidents - by user rights (VD doesn't require elevation to run).

[nji]

Hey shekh

However that's kind of bad news... especially in these times you never know
what you get.

The reason for this post was due of a filter from a somehow strange source.

Still some time ago I already had some other experience which made me feel uneasy:

https://forum.doom9.org/showthread.p...65#post1990265

[shekh]

Originally Posted by nji

Still some time ago I already had some other experience which made me feel uneasy:

https://forum.doom9.org/showthread.p...65#post1990265

Yes I know nothing about MSU but after trying various filters I had impression these are some sort of student projects (nothing good).

[nji]

Recently I wanted to remove an annoying subtitle from a movie.
(A cut of the movie: https://c.gmx.net/@11558428876409458...QqiL0KZkxjFo4g)

I searched and found MSU Subtitle Remover
https://videoprocessing.ai/video_fil...s-removal.html

Trying it (yellow subtitles with a small black border around)
I wasn't successful, no matter what kind of adjustments I did.
There were no or only very small parts of the subtitles found
(even with decreased "Subtitle outline contrast" slider
and/ or increased "Subtitle sensivity" slider).
The present doesn't find anything.
(Anyway I wondered how the filter should work without giving it
explicitely the subtitle color).

Then I tried the previous version (2.2b).
Here you can give the color (in YUV!),
but it doesn't generate any files in working folder.

Has anyone any experiences with it?

Or are there alternative filters for removing subtitles?
Actually I haven't found ANY...

[Emulgator]

InpaintDeLogo

[nji]

Originally Posted by Emulgator

InpaintDeLogo

Thank you.
(I could have bet that with AviSynth it had been done).

But...
that looks extremely complex to me, even if you're common to AviSynth.
Moreover there is a long list of other packets that have to be installed before.
It may be the very last resort, but first of all there may be an origin VD2-filter?
This might be as simple/ usable as the MSU one - but working.


There are possible workarounds, using deshaking filter:

For that there is needed a pre-filter that allows
to specify the main subtitle color/ tolerance, and the area to expand
and fuzzy around that.
As its output the fulfilling area shall
- color-changed (for DeShaker) or
- alpha set (for shekh's repair filter in "lens transform").
DeShaker would have the advantage of intelligent temporal filling.
(But somewhere in my brain I remember that Deshaker doesn't do
with no deshaking (but only ignoring)).
I don't know if there is such a filter somewhere...

(There is - of course - an AviSynth filter: TweakColor.
Still you can't specify parameters for expand and fuzzy area
- which will fail with bordered subtitles).

Any ideas anyone?

[nji]

Take a mp4/avc1 and write it with same codec (but in it's config
change from YUV 4:2:0 to RGB).

Why is that?
I would expect only minor effects, measurable but not so obvious.

https://c.gmx.net/@11558428876409458...QqiL0KZkxjFo4g

[FranceBB]

Hi there,
I've just realized that the last time I updated VirtualDub was like 3 years ago, so I went on to update it again, but the last update I could find is 20 Update 19 from March 20, 2020.
Is this effectively the last version or did I miss something?