Subtitle Edit

[Zetti]

Thanks for new release:
https://github.com/SubtitleEdit/subtitleedit/releases/tag/3.5.15

[Janusz]

@GCRaistlin: 2. The possibility to paste a subtitle from the clipboard. Here's what I mean. Let's consider I have two srt files open in two different copies of SE. In one copy, I do a right-click on a line and select 'Copy as text to clipboard'. In another copy, I do a right-click on a line and select 'Paste from clipboard before'. SE adds a subtitle from the clipboard preserving everything but subtitle number (i. e. start time, duration, text). Why do I need this? It may be useful when syncing subtitles (*) that we downloaded from some site to the ones (**) that we ripped from the disc we want to view with (*) and when subtitles (**) are lacking of some lines that are present in subtitles (*). This way, we could easily copy-and-paste missing lines from (*) to (**), then shift them by applying a delay, then import time codes from (**) to (*).

You have tried: Right click on the line and select <Column / Paste from clipboard ...>
There is what you wrote about.

[GCRaistlin]

Janusz
It is pretty good hidden. Thanks.

[locotus]

Going back to 3.5.14 after noticing that remove leading (...) was remove. Sorry for that.

[Nikse555]

@locotus: "Remove leading ..." is still there with default settings: https://nikse.dk/fix-common-errors.png
But it depends on what "Continuation style" is chosen in "Profile". If you like dash/epllises for both start/end then "Remove leading ..." does not make sense. Read more here: https://github.com/SubtitleEdit/subtitleedit/pull/4108

Thanks for new release:
https://github.com/SubtitleEdit/subtitleedit/releases/tag/3.5.15
Your're welcome.
Yes, 3.5.15 is out :)
Change log: https://raw.githubusercontent.com/SubtitleEdit/subtitleedit/master/Changelog.txt

@GCRaistlin: thx for the sup file, SE should now allow some errors in the file (like for .ts files).
You can also do File -> Import time codes...

[locotus]

@locotus: "Remove leading ..." is still there with default settings: https://nikse.dk/fix-common-errors.png
But it depends on what "Continuation style" is chosen in "Profile". If you like dash/epllises for both start/end then "Remove leading ..." does not make sense. Read more here: https://github.com/SubtitleEdit/subtitleedit/pull/4108

All of the contrary, I only use ellipsis at the end of a dialog that needs to be continue, but never at the beginning
of the continuation because it means losing 3 characters at the start of an already long line and this is what don't make any sense.

Is there any continuation style that only use ellipsis at the end of the line?

[Nikse555]

Is there any continuation style that only use ellipsis at the end of the line?

You can try "Dots (trailing only)": This will also add "..." for continuing lines where it's not present.
There's also "None, dots for pauses (trailing only)".

[locotus]

You can try "Dots (trailing only)": This will also add "..." for continuing lines where it's not present.
There's also "None, dots for pauses (trailing only)".

Thanks for your time but I'd rather stay with 3.5.14.

[tormento]

Yes, 3.5.15 is out
There is a nasty OCR correction in Italian "Fix OCR errors":

It's unusual a «I» is inside a period but when it's the initial letter of a name.

I.e. «I'» is impossibile in italian language, it should always be «l'».

Likewise, «I*» where * is another letter, is very difficult to be found, unless it is the beginning of a personal name or an all capital letter word.

I suggest you to replace every «I» with «l» unless inside a word or at the beginning of a period or paragraph.

The only italian word of two letters that begins with a «I» is «Io», i.e. «I» in english. There are instead definite article «lo», i.e. "male" «the» in english,.usually contracted as «l'» in front of a vowel, that can be a personal pronoun too, i.e. «him» in english. The latter can be contracted as «l'» too.

[Nikse555]

@tormento: Could you give one or two examples? How does it work now and how should it work?
Also, Italian context-menu with letters should now be included in the "OCR character" + "OCR inspect" windows: https://github.com/SubtitleEdit/subtitleedit/releases/download/3.5.15/SubtitleEditBeta.zip

[tormento]

@tormento: Could you give one or two examples? How does it work now and how should it work?
Of course, ASAP I will find some proper ones.
"OCR character" + "OCR inspect" windows
Where is it? I can't find it.

P.S: I suggest you to exclude from merge lines the ones with capital words inside. Include in the list but uncheck them as default.

[tormento]

First example of not corrected «l»:
893
01:28:03,875 --> 01:28:06,500
IL VERO HACHIKO NACQUE
A ODATE, lN GIAPPONE, NEL 1923.

894
01:28:06,583 --> 01:28:08,750
QUANDO lL SUO PADRONE,
IL DOTT. EISABURO UENO,

896
01:28:11,875 --> 01:28:13,667
HACHI TORNÒ lL GIORNO DOPO AD ASPETTARLO

897
01:28:13,750 --> 01:28:16,583
ALLA STAZIONE DI SHIBUYA E
COSÌ FECE PER l SUCCESSIVI 9 ANNI

900
01:28:28,708 --> 01:28:31,667
UNA STATUA lN BRONZO DI HACHIKO
SI TROVA NEL PUNTO ESATTO

901
01:28:31,750 --> 01:28:33,708
IN CUI lL CANE ASPETTAVA lL SUO PADRONE
It should be:
893
01:28:03,875 --> 01:28:06,500
IL VERO HACHIKO NACQUE
A ODATE, IN GIAPPONE, NEL 1923.

894
01:28:06,583 --> 01:28:08,750
QUANDO IL SUO PADRONE,
IL DOTT. EISABURO UENO,

896
01:28:11,875 --> 01:28:13,667
HACHI TORNÒ IL GIORNO DOPO AD ASPETTARLO

897
01:28:13,750 --> 01:28:16,583
ALLA STAZIONE DI SHIBUYA E
COSÌ FECE PER I SUCCESSIVI 9 ANNI

900
01:28:28,708 --> 01:28:31,667
UNA STATUA IN BRONZO DI HACHIKO
SI TROVA NEL PUNTO ESATTO

901
01:28:31,750 --> 01:28:33,708
IN CUI IL CANE ASPETTAVA lL SUO PADRONE

[tormento]

This (https://www.mediafire.com/file/cm2scqov6h1i2ai/The_raid_sup.7z/file) sup gives me problems with Binary image compare.

It seems it is changing color palette and SE can't keep the pace to it. Plus there is graphic corruption when palette change is encountered.

[tormento]

When binary compare OCR subtitles with italic, there are some issues with spaces between words.

Here (https://www.mediafire.com/file/ayayqow1oizpeu1/The_spy_who_sup.7z/file) an example, where it's impossible to have a proper space for normal text and italic on line 9 (ofpower).

The binary compare should apply 2 different "number of pixels is space" to italic and normal text.

[Nikse555]

SE is relying on the "???_OCRFixReplaceList.xml" file to make corrections... sadly there's no "ita_OCRFixReplaceList.xml". I've tried to make one in latest beta: https://github.com/SubtitleEdit/subtitleedit/releases/download/3.5.15/SubtitleEditBeta.zip
Your examples above seem to be fixed I think, but I'm not sure if it breaks something else...
Also, perhaps you could add some fixes/words to it? Perhaps you already have a "ita_OCRFixReplaceList_user.xml" file? You can take a look at the "eng_OCRFixReplaceList.xml" for inspiration.

About the palette, a bit tough, but if you click on "image pre-processing", and select "100" for thresshold, color2white: choose default black, color2remove: choose default white...
Some subtitles uses transparent for "color"... SE cannot handle those.

SE already tries "number-of-pixels" -1 for italic, but I will take a look at the file. Also a lot can be fixed via the OCR fix replace list.

Also, above beta you can customize the "quick-click" letters in the "OCR character" window - search for "OcrAddLetterRow" in Settings.xml. You can also right-click in the text box and get a context menu where Italian letters are available.

[Nikse555]

When binary compare OCR subtitles with italic, there are some issues with spaces between words.

Here (https://www.mediafire.com/file/ayayqow1oizpeu1/The_spy_who_sup.7z/file) an example, where it's impossible to have a proper space for normal text and italic on line 9 (ofpower).

The binary compare should apply 2 different "number of pixels is space" to italic and normal text.

Works fine here with a clean default installation with "pixels is space" set to 9 (edit: up to 11 works fine).

[tormento]

Also, perhaps you could add some fixes/words to it?
I will do my best in spare time.
About the palette, a bit tough
Had a hard time finding where image preprocessing was but now it works, thanks!
Works fine here with a clean default installation with "pixels is space" set to 9 (edit: up to 11 works fine).
Really strange. What setting could influence mine not working?

[Janusz]

This (https://www.mediafire.com/file/cm2scqov6h1i2ai/The_raid_sup.7z/file) sup gives me problems with Binary image compare.

It seems it is changing color palette and SE can't keep the pace to it. Plus there is graphic corruption when palette change is encountered.

First, I did <Save all images with HTML index ...>
Then I did OCR from the index.html file.
The program has read all lines, including those where the colors change.

[GCRaistlin]

Nikse555
Bugs:

If SUP file is open by supplying it as an argument in the command line and got OCRed then after loading the OCR results to the main window SE behaves as if an unchanged SRT file is open: no asterisk in the window title, no prompt to save file on closing. If we won't save the file manually we'll lose our work.
The lack of a point isn't highlighted in 'Compare subtitles' - try to compare these ones (https://mir.cr/0T8EVLL9).
Open a BD SUP file (from the UI), go to, say, 1000th subpic, press 'Start OCR' - 'Stop' - 'Cancel'. SE prompts if we want to discard changes, as expected.
Open the same file (from the UI), don't go anyway, press 'Start OCR' - 'Stop' - 'Cancel'. SE closes without any prompts - NOT as expected.
Open a DVD SUP file (by supplying it as an argument in the command line), go to, say, 1000th subpic, press 'Start OCR' - 'Stop' - 'Cancel'. SE closes without any prompts - NOT as expected.
When we export subpic # 1078 as an image SE offer Image1077.png as a filename.
Open an ASS file (https://mir.cr/AZ3FGZF5). Do 'Copy as text to clipboard' on any line, then 'Column' - 'Paste from clipboard...'. You'll get garbage pasted.


Feature requests:

The ability to define keyboard shortcut for <right-click> - Column - Paste from clipboard...
Make 'Column paste' dialog remember last used radio buttons states. Also, add underscored letters to the radio buttons to make it possible to select it with Alt-<letter>.
The ability to load DVD SUP files from UI.


Why bitmaps that was added to OCR DB during performing OCR of a SUP file don't work when performing OCR of PNG files exported from this SUP file (by 'Save subtitle image as...')? Try to OCR 00262.track_4608.sup (https://mir.cr/0AAWJH75) using my Latin.db (https://mir.cr/0CCIMJS4) from # 1078 - SE won't ask you anything. Then export subpic # 1078, import it and try to OCR it.

You are welcome to import new records from my Latin.db to the supplied one.

[Janusz]

...
Feature requests:
1. The ability to define keyboard shortcut for <right-click> - Column - Paste from clipboard...


Use <Options / Settings / Shortcuts>
In the Search box, type: Col
at the bottom you can also choose "Column, paste". Set the shortcut you want.

[Nikse555]

@GCRaistlin:
>1. After loading the OCR results to the main window SE behaves as if an unchanged SRT file is open
I cannot re-create this... what file format and how do you open the file exactly?

>3. Open a BD SUP...
Yeah, that's a feature. Do any real work and SE will prompt.

>3. The ability to load DVD SUP files from UI...
That works here in main window via File - Open... or drag-n-drop. How/where does it not work?

[GCRaistlin]

Janusz
Thanks again. I saw it but didn't think that this is what I'm looking for. Just wondering why it has a different name here...

[Janusz]

@GCRaistlin

Bugs:
1. After loading the OCR results to the main window SE behaves as if an unchanged SRT file is open: no asterisk in the window title, no prompt to save file on closing. If we won't save the file manually we'll lose our work.

1. I don't know how you are.
For me in the title of the main window after loading the subtitles after OCR the window name changes to: * D: \ full path \ filename.html \ index.srt - SubtitleEdit...
If I am now trying to close the program I get a prompt to save a new file. I don't know why it doesn't work for you.
After saving, the window name indicates where to save the index.srt file without *. If I change anything in the text, the program name will start with * again.

[GCRaistlin]

Janusz
I got it: it happens when SUP file is open by supplying it as an argument in command line, not from UI.

[Nikse555]

@GCRaistlin: Beta updated: https://github.com/SubtitleEdit/subtitleedit/releases/download/3.5.15/SubtitleEditBeta.zip
Fixed dvd sup from cmd line + ass paste + change detection with file+ocr from cmd line + compare issue + image save as number + remember column paste options - hopefully some of that also works you?

[Janusz]

@Nikse555

I don't know how understandable this translator text will be, but I'll try.

In connection with the problem of correct recognition by the OCR systems of the lowercase "L" and the "I",
please explain briefly the principles which follow Subtitle Edit for automatic correction during OCR.

Why i ask?
I start the OCR process with a new, completely empty character base.
My settings: in [OCR auto correction Dictionary] - none, other options unchecked.
No dictionaries in the Dictionaries catalog. Options/Tools [Fix common OCR errors ...] unchecked.
As a result, I get an empty text with spaces recognized according to <No of pixels is space>.
This is correct and as expected.

I add the lower case letter "L" to the character base, options marked as in (1).
As a result, they appear, I don't know if all but definitely recognized lower case "L", but also "I",
which I don't have in the character database. Spaces between words as in (1).
I can agree. At this stage I accept "I", because in Polish lonely "I" occurs so it is OK.
But why does the program change the lowercase "L" to the "I" at the beginning of words
since they do not yet know these words. It looks like this: [I I**** *********].
The first "I" - ok. The second may well be the lower case letter "L".
I consider this a serious mistake.
Such a conversion could take place only after recognizing the entire word and after checking
in the selected dictionary that such a conversion would not cause an error.
From what you can see - checking in the dictionary is missing or not working as it should.
I omit the fact that now the dictionary is off because every word is a mistake.

Another OCR attempt.
a. I remove the only lower case letter "L" from the character base. Checking - the character database is empty. Start OCR - result as in (1) - OK.
b. Another OCR attempt. The character base is the same. However, the small "L", previously removed, is still there. I do not know why?
Did the program not save the changes permanently? OCR result as for (2) with all errors ("I").
I tried in various ways to get rid of the stubborn letter from the base,
but you can't delete a character if it is the only character in the database.
The changes only apply in the current program session.
After closing and restarting the program, everything (last letter) returns to the state it was before.
I think it's a mistake. Or maybe it should work like that? I do not know.

Finally, my suggestion to consider in the distant or near future.
When automating the OCR process, give the opportunity to use (set) a second dictionary. Subtitles are usually a translation of one language into another.
However, proper names, first names, last names etc. which do not have equivalents in a given language are often not translated.
Using a second language outside the main language - will generate fewer errors, making the process more intelligent.
We don't always create or correct subtitles in one language.

If the translation is not understandable enough, I am sorry.
I wanted to help You and myself.

[tormento]

@GCRaistlin: Beta updated
I really can't find a pixel space to binary OCR the italic part of this (https://www.mediafire.com/file/hx912s8z3ftw56b/500_sup.7z/file) sup. No problems at all with SupRip.

Can you please tell me if you can and what parameters are you using?

[GCRaistlin]

Fixed dvd sup from cmd line
Not sure what you mean. The other fixes are confirmed, thank you.


>3. Open a BD SUP...
Yeah, that's a feature. Do any real work and SE will prompt.

I did the real work actually: I opened a BD SUP file, started OCR from the beginning, then aborted it at some point. There is recognized data, but if I press Cancel SE discards it silently. Though if I do the same steps but start OCR from # 1000 SE asks me about discarding. Why the behavior is different?
My logic is pretty simple: if there is any recognized data (even just one char) SE should ask about discarding if the user press Cancel or tries to close the window.


>3. The ability to load DVD SUP files from UI...
That works here in main window via File - Open... or drag-n-drop. How/where does it not work?
Yeah, this works. I should say though that it is completely unintuitive: we should use 'File - Import/OCR' for BD SUP files but 'File - Open' for DVD SUP files.

Bug: if we open a DVD SUP from the command line and then press Cancel in 'Import/OCR' dialog SE remains open.

[Nikse555]

@GCRaistlin:
"Fixed dvd sup from cmd line" is about the "*" in the title bar.

OK, OCR window should now prompt for save changes if anything has been added.

File - Import/OCR' for BD SUP will now also allow dvd sup. I just always use File -> Open...

>Bug: if we open a DVD SUP from the command line and then press Cancel in 'Import/OCR' dialog SE remains open.
Thx, should now hopefully be fixed.

Latest beta updated: https://github.com/SubtitleEdit/subtitleedit/releases/download/3.5.15/SubtitleEditBeta.zip



I really can't find a pixel space to binary OCR the italic part of this (https://www.mediafire.com/file/hx912s8z3ftw56b/500_sup.7z/file) sup. No problems at all with SupRip.

Can you please tell me if you can and what parameters are you using?

Sorry, SE cannot handle that file via "Binary image compare".

[tormento]

Sorry, SE cannot handle that file via "Binary image compare".
SupRip tilt OCR frames according to character vertical lines.

Isn't possible to implement something like that into SE?

[tormento]

@Nikse

This (https://www.mediafire.com/file/xtf28xmnndnc8bf/apes_sup.7z/file) too has problem with Italic. Perhaps there was some regression at some point, because almost all titles I am doing OCR are "cursed".

Can you fix the binary compare engine, regarding Italic? You are telling that it's easily fixed by OCR error rules but I can't find any universally suitable.

[GCRaistlin]

File - Import/OCR' for BD SUP will now also allow dvd sup. I just always use File -> Open...
Maybe it's better then to remove 'Import/OCR VobSub' and 'Import/OCR BluRay (.sup)' menu items since such files can be opened via File - Open? If you think it's too much then please at least rename 'Import/OCR BluRay (.sup)' to simple 'Import/OCR .sup subtitle file'.

Feature requests:

Ability to set font properties for Spell Check dialog window.
Ability to run Spell Checking from within 'Import/OCR' window after performing OCR - with the possibility to select the subpic with the currently unknown word. Why don't I perform spell checking during the OCR? Because I prefer adding a better match than using 'Fix common OCR errors' or spell check (it's better to prevent errors than to fix them). I consider spell check as a tool that helps finding errors that were missed first during the OCR and then during visual check.


A lot of mistakes in word boundaries detection could be avoided if the vertical lines of a character, if any, were taken into account in the first place. Therefore we need a new setting - 'No of pixels from/to vertical line is space'; this setting takes precedence over simple 'No of pixels is space'. Examples:

"of July" in this subpic:
https://i111.fastpic.ru/thumb/2020/0505/71/71c4ab87d968ccea6422ddf2ccdb0871.jpeg (https://fastpic.ru/view/111/2020/0505/71c4ab87d968ccea6422ddf2ccdb0871.png.html)
is now being recognized as "ofJ uly" - two errors in once (with 'No of pixel is space' setting of 9 which seems to be the most reliable choice for all subpics in the sup file). The modified algorithm could process it as follows:
For "f", there is 27 running dots with the same X position from the right boundary of 47 total "vertical" dots; hence, we have a vertical line from the right side here.
For "J", there is 46 running dots with the same X position from the left boundary of 56 total "vertical" dots; hence, we have a vertical line from the left side here.
There are 26 pixels between these two vertical lines while there are only 6 pixels between the most right dot of "f" and the most left dot of "J".
There are 10 pixels between "J" and "u".
"If you" in this subpic:
https://i111.fastpic.ru/thumb/2020/0505/41/9f8eb4b995d0e5d8e8ba83ccb473c041.jpeg (https://fastpic.ru/view/111/2020/0505/9f8eb4b995d0e5d8e8ba83ccb473c041.png.html)
is now being recognized as "Ifyou" (with 'No of pixel is space' setting of 9). The modified algorithm could process it as follows:
We have the same case with "f" as above.
With "y", we don't have a vertical line from the left side as there's no enough quantity of running dots with the same X position from the left boundary.
Hence, we take into account the right vertical line of "f" and the most left dot of "y" when counting pixels between these characters: 19 pixels.

With 'No of pixels from/to vertical line is space' setting of 19 and 'No of pixel is space' setting of 10 both problematic substrings above would be recognized properly.

[GCRaistlin]

Feature request: use different fonts for list view and text boxes. For text boxes, Courier New is sometimes a better choice than Tahoma as it clearly shows the difference between a double quote and doubled apostrophe (OCR error). But list view looks ugly with Courier.

[GCRaistlin]

There's some incompatibility with Ditto (https://sourceforge.net/projects/ditto-cp/files/?source=navbar), the clipboard manager:

Download and install Ditto.
Import the .reg file:

[HKEY_CURRENT_USER\Software\Ditto]
"DittoHotKey"=dword:00000857

It sets Win+W keyboard shortcut to call Ditto.
Run Ditto, copy some text to the clipboard.
Open SE, place the cursor to the text box.
Press Win+W, select the text in the list, press Enter.

The text is pasted. Now you are unable to edit the text in the text box (though you can paste there).

[tormento]

If possible, do no automatically select remove line break when the next line starts with a capital letter.

[varekai]

Just wanted to say thanks for Subtitle Edit, couldn't do without it!
I'm not much of a bugtester (there seems to be others...;) as SE works perfectly for me!
So... I feel there's no need for changing the GUI.

[varekai]

@GCRaistlin

Please, consider another image hosting...:devil:

[GCRaistlin]

varekai
You are welcome to offer another one if it is as handy as the current one is.

[varekai]

This is how I would do it... (https://i.imgur.com/pmU53W7.jpg)

[Nikse555]

@GCRaistlin: Regarding Ditto, I have seen the error like 1 time in 100 pastes and I'm afraid I've no idea what's wrong or how to fix it. Ideas/fixes are welcome.

@tormento: Writing a new image-to-letter-splitter and integrating it into SE is a lot of work - could easily take 14+ days full time.
Is SupRip open source?
Edit: Thx for the sample files with italic! Surely uses more tilt than I've seen before.

[Nikse555]

Just wanted to say thanks for Subtitle Edit, couldn't do without it!


You're welcome, and thx :)

[GCRaistlin]

Regarding Ditto, I have seen the error like 1 time in 100 pastes

That's strange. I have the lock error every time I'm pasting from the Ditto list.

[GCRaistlin]

This is how I would do it... (https://i.imgur.com/pmU53W7.jpg)

Without the preview? A good change, indeed.

[jlw_4049]

A couple bugs. 'I' wants to be |
When using the latest version, both the original Tesseract 3x and 5x on certain parts of the PGS subtitle file is zoomed in to far for me or the program to read what it is. Other then those problems the newer version seems to perform better.

[varekai]

Without the preview? A good change, indeed.
(drumroll)
https://i.imgur.com/3DcrNTy.png
OKEYYY!
https://i.imgur.com/NeOY2LN.jpg
Are you just playing stupid or are you really... ;)

Edit
...clarification...
https://i111.fastpic.ru/thumb/2020/0505/71/71c4ab87d968ccea6422ddf2ccdb0871.jpeg

https://i111.fastpic.ru/thumb/2020/0505/41/9f8eb4b995d0e5d8e8ba83ccb473c041.jpeg

[Nikse555]

A couple bugs. 'I' wants to be |
Could you upload a sample where this happens?
SE tries to fix this via the eng_OcrFixReplaceList.xml file...

[jlw_4049]

Could you upload a sample where this happens?

SE tries to fix this via the eng_OcrFixReplaceList.xml file...I'll send the sample first thing tomorrow when I get up with the line(s) that it does it on.

Thanks!

Edit: Also adding the OCR corrected words in this format fixed the I issue.

-| to I

Sent from my Pixel 3a using Tapatalk

[GCRaistlin]

Nikse555
Regarding Ditto again. The issue seems to be worse that I thought: it is enough to just call the Ditto list on the screen (Win+W) having SE active to get the lock error. But it is only the main screen text box that gets locked - text boxes in Options and 'Import/OCR' window are immune to this.

[Nikse555]

Nikse555
Regarding Ditto again. The issue seems to be worse that I thought: it is enough to just call the Ditto list on the screen (Win+W) having SE active to get the lock error. But it is only the main screen text box that gets locked - text boxes in Options and 'Import/OCR' window are immune to this.

OK, how is latest beta?
https://github.com/SubtitleEdit/subtitleedit/releases/download/3.5.15/SubtitleEditBeta.zip

[GCRaistlin]

Nikse555
Seems to be fixed, thanks!