InpaintDelogo - advanced logo removal & hardcoded subtitles extraction

[VoodooFX]

Don't use settings you don't understand, your script is complete nonsense.

PHP Code:

LWLibAvVideoSource("D:\Professor.mp4")

# Back to original resolution:
Crop(160,0,-160,0).Spline16Resize(640,480)

# Remove this after Automask run:
Trim(90,40480)

InpaintDelogo(mask="D:\mask.bmp", Loc="336,420,-12,-8", Automask=0, Mode="Inpaint", Analyze=3) 


Before:


After:

[Spiny Norman]

If someone could give some beginner-level advice on subtitle extraction, I'd be very grateful.

I very often come across old SD videos that have burnt-in subtitles that may still be useful. Those subtitles are often wanted by someone. So I set out from scratch to see if I could (learn how to) extract them.

So to use an example: What settings would work best for example to extract from this video?

Let's say https://www.youtube.com/watch?v=bJKq7SVNfkM for the simple reason that it's SD but has very legible subtitles. (I have some that are in worse quality but that's for later.)

LWLibavVideoSource("C:\Swtools\Backanterna.mp4")
InpaintDelogo(Loc="94,368,280,86", Show=4, DynMask=4, DynTune=160, DynMask4H=70, Extract=1, ImgDir="C:\Swtools")

Thanks to VoodooFX I installed the required stuff including plugins and I can manage the 'loc' part and I'm extracting png files, like the attached example.

But I could probably get better output, right? I've seen mentioned DynMask=5 and I guess in some cases I should specify something with DynColor with a hexadecimal value. (For example if the colours are different such, like on the arte channel: https://www.youtube.com/watch?v=vNwFwdEtpy0)

How can I improve the script to get better results in cases like these?
(Most of this stuff is new for me, so it's possible an exact argument has to be spelled out for me. But anything that gets me further on the rails I'll very much appreciate!)

[VoodooFX]

Originally Posted by Spiny Norman

If someone could give some beginner-level advice on subtitle extraction, I'd be very grateful.

From your post I can tell that you are not doing anything yet, so read the manual first.

Both videos are low quality, first learn the subs extraction with the high quality videos.

[Spiny Norman]

Out of interest: Is there a specific reason to start with HD?

Meanwhile, if anyone has suggestions how to expand my first attempt above to get better results, I'm all ears.

[VoodooFX]

[Spiny Norman]

So I have the greatest respect for your programming skills, but nevertheless I'd be interested in a more specific reference, or to be given/pointed to an example. Even "see post ##" would already be something. As before, you (the OP) can pass if it doesn't interest you.

As said, a lot of discussion concerns delogoing which is (probably?) irrelevant.
The manual describes the functions but that doesn't directly lead to their application from my POV. (On a theoretical level, all information is already somewhere, but that is not how learning works. Otherwise all schoold etc. could close for good. We ALL (yes, OP included) have at one time or another asked questions that we could have found somewhere - in retrospect.)

Let me change the question completely: Has anyone tried this (sub. extraction) on old VHS sourced material? If so, how did it go?

[VoodooFX]

Originally Posted by Spiny Norman

...to be given/pointed to an example.

You already posted an example yourself, there -> https://forum.doom9.org/showthread.php?p=2007676
Which is a copy/paste from somewhere, with only Loc adjusted.

Originally Posted by Spiny Norman

The manual describes the functions but that doesn't directly lead to their application from my POV.

The manual directly leads to application, it's oxymoron to think otherwise.

Originally Posted by Spiny Norman

Let me change the question completely: Has anyone tried this (sub. extraction) on old VHS sourced material? If so, how did it go?

Yes. It went well.

PS: No need to post same things on the three different forums.

[Spiny Norman]

That's the whole point, isn't it: The example I gave is probably not the best possible outcome. That wouldn't make sense given your earlier demonstrations of results.

Originally Posted by VoodooFX

No need to post same things on the three different forums.

Yes there is. On the other hand you definitely need not reply on all three. You don't even need to reply on one if you don't want.

(You're making it very hard. You yourself have on occasion also said something wrong or asked for information you could have found. How would you have liked to be put down for something like that?)

[VoodooFX]

Originally Posted by Spiny Norman

That's the whole point, isn't it: The example I gave is probably not the best possible outcome.

It's crazy to expect that copy/paste of some random script example would have any meaningful output on various random videos.
You need to adjust the script example to video. To be able to do that you need to read the manual.

"The whole point": What prevents you reading the manual?

Originally Posted by Spiny Norman

You yourself have on occasion also said something wrong or asked for information you could have found. How would you have liked to be put down for something like that?

You are implying here some nonsense, I've no idea why you post these fantasies.

[Spiny Norman]

Originally Posted by VoodooFX

You are implying here some nonsense, I've no idea why you post these fantasies.

Originally Posted by VoodooFX

What prevents you reading the manual?

When you say it like that, it seems like a polite way to say F-- off.
But I can give a completely serious answer, if it is a real question. Outside the technical dimension there might after all be things that you didn't think of yet.

[VoodooFX]

This topic is outside of your dimension.
Go see the doctors, good luck.

[Spiny Norman]

Well, I'm sorry you feel that way. There might be more to it then "RTFM", but as said, you're free to pass if explaining with examples is not up your alley - no-one could expect you to answer ALL questions.

Thank you for the help so far and see you in the waiting room then.

[Guilbert]

Hi,

I sent you a video, VoodooFX.
I think the parameters are good, I have much better results than with VideoSubFinder, except that more than 30 subtitles are missing compared to VSF (https://www.subdiff.com/view-EgsS7cG5ej.html), and I don't understand why.
I tried to play with SubMinDur, SubMinLuma and SubSusDur with no luck...

Here is my parameters:

InpaintDelogo(Loc="400,800,-400,-40", Show=4, DynMask=3, Dyn3Seq=24, DynTune=190, DynTEdge=1, DynTEdgeThY=20, DynTEdgeInf=2, DynLocUp=2, CorrTh=0.7)

Can you explain to me what's wrong, please?

Thank you for your help.

[VoodooFX]

Originally Posted by Guilbert

Can you explain to me what's wrong, please?

Hi, "Dyn3Seq=24" - it means that subs shorter than 1 sec [your video is ~24 fps] will disappear from a mask, so probably all those 30 subs are shorter than 1 sec.
Read about Dyn3Seq in the manual.

And I would use such settings:

InpaintDelogo(Loc="400,800,-400,-40", Show=4, DynMask=3, Dyn3Seq=18, DynTune=210, DynTEdge=1, DynTEdgeThY=15, DynTEdgeInf=4, DynLocUp=2, CorrTh=0.8)