To improve automation support: allow selecting one track specifically, and support providing the exact name for the outputted subtitle file

[varenc]

I love this project! Thanks for making it. The Vision framework has great OCR and glad to see it getting put to good use.

I'd like to use macSubtitleOCR to generate subtitles en-masse for some old 1960s media. Right now there's two things that make this tricky:

• macSubtitleOCR always OCRs all of the image subtitles on the input media. Specifying -l en to select just the image I want doesn't change its output. I assume this is just a signal to the OCR framework about what languages to expect.
  • Solution: support a '--sub-id` parameter or something similar that lets me specify the exact index of the subtitle I care about. Usually specifying a language would be enough, but sometimes media incorrectly doesn't label its languages, and sometimes the same language has multiple subtitle tracks ("Foreign Parts Only", "Hearing Impared", etc)
• macSubtitleOCR only lets you specify an output directory and the outputted file names are always track_<INDEX>.srt. This makes automation more tiresome.
  • Solution: when only one inputted subtitle track is selected, allow me to specify the output filename exactly.

Putting both of these together would let me do something like this:

macSubtitleOCR /path/to/input.mkv --sub-id 3 /path/to/index.en.srt

The output names are annoying, but definitely fixable with slightly more complicated automation. The bigger annoyance is that it always processes all subtitles, meaning that for media with 20 different languages it's taking about 19 times longer than it needs to be for what I care about.

I make try to put together a PR with these little issues, so just putting it on your radar.

Thanks again for making this!

[ecdye]

Thanks for the suggestions, I would be happy to review a PR with options like this.

[timj]

I have a python wrapper that analyzes all the tracks using mkvmerge, then extracts the subtitles (.idx or sup or whatever) and then runs the subtitle OCR on each of those in turn and gives them a reasonable name -- I have more control over what happens when I do that but it does rely on mkvtoolnix being installed.