pgsrip icon indicating copy to clipboard operation
pgsrip copied to clipboard

Problems with stylized PGS subs on some files

Open outlyer opened this issue 10 months ago • 1 comments

I've had very good results with most of the files I've converted, but I have noticed the OCR seems to be particularly bad in some situations. I've narrowed this down to the specific way that the PGS subtitles are styled and how they are processed.

With a particularly styled file, the OCR is really inaccurate. I used the --keep-temp-files to see the files, and it looks like the text is inverted and placed on a black background, but the way these particular subtitles are formatted, they show up as a mostly black file.

Here is a normal file:

english srt-1539-psm6-NEURAL-65

and here is an example of the issue:

Blue Collar example

The second example has a border around the font which seems to be the cause of the issues.

outlyer avatar Apr 10 '24 22:04 outlyer