AviSubDetector v0.6.0.5

Shalcker · 11th February 2005, 04:54

AVISubDetector is a tool for extraction and recognition of hard-burned subtitles from various video sources.

Latest version is available here:
http://animeburg.omake.ru/download/u...or_english.zip
http://animeburg.omake.ru/download/u...or_english.rar

Source Code (GPL):
http://animeburg.omake.ru/download/u...or_0605src.zip

Older thread about this program is available here:
http://forum.doom9.org/showthread.php?s=&threadid=56062

gubi-gubi · 14th February 2005, 20:55

Whats the changes?

Shalcker · 15th February 2005, 02:52

Quote:

Originally posted by gubi-gubi
Whats the changes?

Only a few minor changes from 0.6.0.2.

A bit of localization support (See "File\Language").
Few visualization fixes (there was several >= while actual program used >).
Option to capture subtitle images with full frame width - since sometimes program fails to properly set subtitle edges and some part of subtitle is cut.

..and now you can download sources too

seewen · 15th February 2005, 02:56

This tool is EXCELLENT.
I'll try it at once. Thanks.

seewen · 15th February 2005, 13:37

There's no readme file at all with this tool ?

Shalcker · 15th February 2005, 14:45

Quote:

Originally posted by seewen
There's no readme file at all with this tool ?

Yes, there is no readme. There was one but it pointed to aforementioned older thread. There was even older one with basic settings but that one is also explained in older thread.

Most settings have "hints" that should appear fairly fast as you hover over them. It's hard for me to explain them better then i've done in that thread - but if you still have questions about interface you should ask them and hopefully i'll be able to make it clearer.

gubi-gubi · 16th February 2005, 01:44

This is truley an awesome tool. It does deserve a proper tutorial though... Anyone fancy it?

johner23 · 16th February 2005, 02:00

I really enjoy your program.

Very useful to "rip" subtitles from my old VHS collection ( currenty stored into AVI ou MPEG-2 formats in my dvd collection ).

The only thing that worry me is that AVISubDetector is very fully about features and become a little bit "hard" to manage for beginners.

If someone wanna try to create a General Guide for it, it will be very useful.

PS: do you code it using Assembly, C/ C++ or wich language ? BTW, congratulations for these very handy and nice software.

Best Regards.

johner

Shalcker · 16th February 2005, 03:26

Quote:

PS: do you code it using Assembly, C/ C++ or wich language ?

Program was created in Delphi (Pascal) - since in the beginning it was more like a test of one simple idea, and the faster i could get working thing, the better... and in the end it was too big for major rewrite

I can probably manage to move some functions into DLL and write those in C/C++ or assembler for speed.

johner23 · 21st February 2005, 00:39

Hi, Shalcker.

I'm trying to extract a hardcoded subtitle from an avi.

But the problem is that stupid subtitles are very white and pale. And the movie is very clear too. Many "lights" around the scenes.

When I did the same with "yellow-golden" colored subtitles, in other movies, no problem.

But now, with those white and pale ones, ASD can't recogize them.

What should I do ? Any setting trick ?

If you could help, I will be very grateful.

Thanks in advance.

johner

Shalcker · 21st February 2005, 05:08

Quote:

Originally posted by johner23
But now, with those white and pale ones, ASD can't recogize them.

What should I do ? Any setting trick ?

There is too many variations of video, "description" in most cases is not enough. Can you post image with example somewhere and give link here?

Mouse · 21st February 2005, 16:33

Shalcker, GREAT Software!

Question;

Can you possible add support for Swedish chars ?
(we use them stupied dots over a and o )

- look like this ;

(hope this translates OK ?)

å Uppercase = Å
ä Uppercase = Ä
ö Uppercase = Ö

I'll add an image, just in case.

Mouse · 21st February 2005, 17:07

Image...

svcdprayer · 27th February 2005, 14:40

Hello! Anyone can tell me if theres anyone wrote a guide how to properly use this great software?

Thanks

Congratulations to the author of the software!

MickRC · 15th March 2005, 16:52

OCR is freezing up after some time. Could u please fix that?
Great program, anyway!

johner23 · 15th March 2005, 19:25

Hi, my friend Shalcker.

A very great program, no doubt.

Do you remember me ? I sent you some pictures from white-pale subtitle hard burned in a movie.

ASD can't read them properly ( don't recognize them )and I sent you some pics to analise it.

If you get time, progress and patience for solve those problems, let me know, ok ?

I repeat: when I use an yellow/golden subs to rip, ASD solve the task without any problem. But the blue/white pales ones, it doesn't.

Thank you very much.

johner

Shalcker · 15th March 2005, 20:50

Quote:

Originally posted by johner23
If you get time, progress and patience for solve those problems, let me know, ok ?

Well, just lower Drop settings to 60 instead of default 100... since there is no outline around these subtitles AND they are from low-resolution video. Either that or changing "shift" from 1 to 3. Program should be able to detect them then (or at least it does for samples you provided to me).

MickRC · 16th March 2005, 01:40

I'm having a problem with subtitles that are in interlace.
Sometimes, first line of subtitle that appears is lower line. Upper is unrecognizable, presented with those interlace lines. On the next frame, both lines are complete. So, I had to reduce "Skip Change if Distance is Less than" to 0, to be able to get both lines. Now, program detects change and result is like this:

4
00:25:47,480 --> 00:25:47,520
starac, morat ces me slusati.

5
00:25:47,520 --> 00:25:52,320
Ako kanis postati smezurani
starac, morat ces me slusati.

As you can see, time difference between subs 4 and 5 is only one frame.
Later I have to manually delete subs like 4 since they are not required. I was just thinking, something like option "Start OCR X Frames After Change is Detected" would solve the problem.

Other problem is with local East European letters. We have S with \/ on top of it. Couldn't find a way to configure S and \/ as one letter.

Shalcker · 16th March 2005, 02:14

Quote:

Originally posted by MickRC
As you can see, time difference between subs 4 and 5 is only one frame.
Later I have to manually delete subs like 4 since they are not required. I was just thinking, something like option "Start OCR X Frames After Change is Detected" would solve the problem.

Well, you can also deinterlace using avisynth before feeding such source to AVISubDetector, if i get you right... or tweak detection settings if by "interlaced" you mean something different...

Quote:

Other problem is with local East European letters. We have S with \/ on top of it. Couldn't find a way to configure S and \/ as one letter.

"Areas" are united into one symbol when they are no farther then "OCR/Main/Symbol Rects/Max Diacritic Distance" vertically. Increase or decrease it to match your font size. Default 32 is around 8 real pixels, so it should be sufficient unless there are really big fonts used.

MickRC · 16th March 2005, 02:15

I've been experimenting with your cool program lately, and would like to share my experience.
Source material is capture from digital Satellite TV, mpeg2, picture is sharp and clear, and so is the subtitle. Letters are white, on grey background. Input video resolution is 704x576.
First, I filter the picture in VirtualDub.
-resize to 4:3 aspect ratio
-mask unncessary areas with grey, using logoaway filter with solid fill option. Chosen color is 0x373737, color close to background of letters.
(First logoaway filter from top of the screen to top of first line of subtitles, second logoaway filter from bottom of 1st line of subs to top of 2nd line of subs, third from bottom of 2nd line of subs to bottom of the screen. Fourth and fifth mask left and right of where subs appear.)
Why? Well, I'm interrested only in subtitles, and that is what I want AVISubDetector to see. It gives OCR an easier job, less artefacts and resulting ripped subtitle has less garbage in it.

Then, using frame server option in VirtualDub, I load filtered video in AVISubDetector.
Options that are different than default are:
-settings-crop
-Drop Values-Shift 4
-Blocks-Value 780
-DLC 17
-L/RMB 60
-OCR-Main-Clearing settings-Alternative base, unchecked all three of Remove open Area From Text Mask, and (020) Match Text Mask to Outline Mask. (I guess that in theory this is used for removing garbage from the picture, but in practice they remove normal letters as well. So, I don't use them, sry.)
-OCR-Main-OCR Setting-Required Similarity 95%.

Once I defined all symbols, I started to use REMOVE Unknown-All. And Required Similarity to 90%.

And now it is pure pleasure to watch your great program in action!

11th February 2005, 04:54	#1 \| Link
Shalcker Programmer Join Date: Mar 2003 Location: Ekaterinburg, Russia Posts: 112	AviSubDetector v0.6.0.5 AVISubDetector is a tool for extraction and recognition of hard-burned subtitles from various video sources. Latest version is available here: http://animeburg.omake.ru/download/u...or_english.zip http://animeburg.omake.ru/download/u...or_english.rar Source Code (GPL): http://animeburg.omake.ru/download/u...or_0605src.zip Older thread about this program is available here: http://forum.doom9.org/showthread.php?s=&threadid=56062 __________________ AVISubDetector - tool for extraction of hard-burned subtitles Last edited by Shalcker; 8th May 2005 at 18:29.

21st February 2005, 16:33	#12 \| Link
Mouse Registered User Join Date: Jun 2002 Posts: 57	Shalcker, GREAT Software! Question; Can you possible add support for Swedish chars ? (we use them stupied dots over a and o ) - look like this ; (hope this translates OK ?) å Uppercase = Å ä Uppercase = Ä ö Uppercase = Ö I'll add an image, just in case. Last edited by Mouse; 21st February 2005 at 17:06.

21st February 2005, 17:07	#13 \| Link
Mouse Registered User Join Date: Jun 2002 Posts: 57	Image... Attached Images

15th March 2005, 19:25	#16 \| Link
johner23 Registered User Join Date: Jul 2003 Location: Brazil Posts: 197	Hi, my friend Shalcker. A very great program, no doubt. Do you remember me ? I sent you some pictures from white-pale subtitle hard burned in a movie. ASD can't read them properly ( don't recognize them )and I sent you some pics to analise it. If you get time, progress and patience for solve those problems, let me know, ok ? I repeat: when I use an yellow/golden subs to rip, ASD solve the task without any problem. But the blue/white pales ones, it doesn't. Thank you very much. johner Last edited by johner23; 15th March 2005 at 19:33.

Thread Tools
Show Printable Version Email this Page
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

14th February 2005, 20:55	#2 \| Link
gubi-gubi Registered User Join Date: Jul 2004 Posts: 29	Whats the changes?

15th February 2005, 02:56	#4 \| Link
seewen Registered User Join Date: Jun 2002 Posts: 250	This tool is EXCELLENT. I'll try it at once. Thanks.

15th February 2005, 13:37	#5 \| Link
seewen Registered User Join Date: Jun 2002 Posts: 250	There's no readme file at all with this tool ?

16th February 2005, 01:44	#7 \| Link
gubi-gubi Registered User Join Date: Jul 2004 Posts: 29	This is truley an awesome tool. It does deserve a proper tutorial though... Anyone fancy it?

16th February 2005, 02:00	#8 \| Link
johner23 Registered User Join Date: Jul 2003 Location: Brazil Posts: 197	I really enjoy your program. Very useful to "rip" subtitles from my old VHS collection ( currenty stored into AVI ou MPEG-2 formats in my dvd collection ). The only thing that worry me is that AVISubDetector is very fully about features and become a little bit "hard" to manage for beginners. If someone wanna try to create a General Guide for it, it will be very useful. PS: do you code it using Assembly, C/ C++ or wich language ? BTW, congratulations for these very handy and nice software. Best Regards. johner

21st February 2005, 00:39	#10 \| Link
johner23 Registered User Join Date: Jul 2003 Location: Brazil Posts: 197	Hi, Shalcker. I'm trying to extract a hardcoded subtitle from an avi. But the problem is that stupid subtitles are very white and pale. And the movie is very clear too. Many "lights" around the scenes. When I did the same with "yellow-golden" colored subtitles, in other movies, no problem. But now, with those white and pale ones, ASD can't recogize them. What should I do ? Any setting trick ? If you could help, I will be very grateful. Thanks in advance. johner

27th February 2005, 14:40	#14 \| Link
svcdprayer Registered User Join Date: May 2002 Posts: 192	Hello! Anyone can tell me if theres anyone wrote a guide how to properly use this great software? Thanks Congratulations to the author of the software!

15th March 2005, 16:52	#15 \| Link
MickRC Registered User Join Date: Mar 2005 Posts: 9	OCR is freezing up after some time. Could u please fix that? Great program, anyway!

16th March 2005, 01:40	#18 \| Link
MickRC Registered User Join Date: Mar 2005 Posts: 9	I'm having a problem with subtitles that are in interlace. Sometimes, first line of subtitle that appears is lower line. Upper is unrecognizable, presented with those interlace lines. On the next frame, both lines are complete. So, I had to reduce "Skip Change if Distance is Less than" to 0, to be able to get both lines. Now, program detects change and result is like this: 4 00:25:47,480 --> 00:25:47,520 starac, morat ces me slusati. 5 00:25:47,520 --> 00:25:52,320 Ako kanis postati smezurani starac, morat ces me slusati. As you can see, time difference between subs 4 and 5 is only one frame. Later I have to manually delete subs like 4 since they are not required. I was just thinking, something like option "Start OCR X Frames After Change is Detected" would solve the problem. Other problem is with local East European letters. We have S with \/ on top of it. Couldn't find a way to configure S and \/ as one letter.

16th March 2005, 02:15	#20 \| Link
MickRC Registered User Join Date: Mar 2005 Posts: 9	I've been experimenting with your cool program lately, and would like to share my experience. Source material is capture from digital Satellite TV, mpeg2, picture is sharp and clear, and so is the subtitle. Letters are white, on grey background. Input video resolution is 704x576. First, I filter the picture in VirtualDub. -resize to 4:3 aspect ratio -mask unncessary areas with grey, using logoaway filter with solid fill option. Chosen color is 0x373737, color close to background of letters. (First logoaway filter from top of the screen to top of first line of subtitles, second logoaway filter from bottom of 1st line of subs to top of 2nd line of subs, third from bottom of 2nd line of subs to bottom of the screen. Fourth and fifth mask left and right of where subs appear.) Why? Well, I'm interrested only in subtitles, and that is what I want AVISubDetector to see. It gives OCR an easier job, less artefacts and resulting ripped subtitle has less garbage in it. Then, using frame server option in VirtualDub, I load filtered video in AVISubDetector. Options that are different than default are: -settings-crop -Drop Values-Shift 4 -Blocks-Value 780 -DLC 17 -L/RMB 60 -OCR-Main-Clearing settings-Alternative base, unchecked all three of Remove open Area From Text Mask, and (020) Match Text Mask to Outline Mask. (I guess that in theory this is used for removing garbage from the picture, but in practice they remove normal letters as well. So, I don't use them, sry.) -OCR-Main-OCR Setting-Required Similarity 95%. Once I defined all symbols, I started to use REMOVE Unknown-All. And Required Similarity to 90%. And now it is pure pleasure to watch your great program in action!