Fix OCR errors option ? about pgstosrt HOT 5 OPEN

@Tentacule Just bumping this one.

I've run a bit of a test with the latest version of PgsToSrt and Subtitle Edit w/ Tesseract 5.3.3
I can't seem to replicate the accuracy of Subtitle Edit using PgsToSrt, even with Fix OCR errors unchecked. Some of the conversion, which should be pretty basic, come out as gibberish (see screenshot).

Command I'm using for PgsToSrt is:
dotnet PgsToSrt.dll --input "file.mkv" --tracklanguage eng --tesseractdata "C:\Program Files\Tesseract-OCR\tessdata" --tesseractlanguage eng

Any ideas here?

from pgstosrt.

I won't add "Fix OCR errors" for now because this functionality is not included in LibSE.

I have done some tests, it looks like an issue on windows, it's working fine when run on linux. I'll investigate.

from pgstosrt.

I just flicked you an email.

I don't think "Fix OCR errors" will make a difference anyway as I had it disabled in SE (see first screenshot) and it still converted the PGS subs almost perfectly. Issue is something else.

Thanks for looking into it.

from pgstosrt.

There was an isssue in windows Tesseract dll, I tried another one and it looks good now.

Here is a new release with this change: PgsToStr-1.4.5.zip

from pgstosrt.

Can confirm that 1.4.5 fixes it. Does a much better job at conversion with no random gibberish to be seen.
Tested with eng.traineddata from:

• https://github.com/tesseract-ocr/tessdata_best
• https://github.com/tesseract-ocr/tessdata

Command:

dotnet "PgsToSrt-1.4.5\\PgsToSrt.dll" --input "file.mkv" --tracklanguage eng --tesseractdata "C:\\Program Files\\Tesseract-OCR\\tessdata" --tesseractlanguage eng

dotnet "PgsToSrt-1.4.5\\PgsToSrt.dll" --input "file.mkv" --tracklanguage eng --tesseractdata "C:\\Program Files\\Tesseract-OCR\\tessdata_best" --tesseractlanguage eng

On the test I ran with the english subtitles for the movie Blade, using tessdata_best took just under 4 minutes with mostly perfect results, while tessdata took 1 minute and had only a few very minor mistakes e.g. capital "I" instead of "i" etc.

from pgstosrt.