Generating English subtitles for International TV

November 15, 2008 2:52 AM

I want to auto-generate English subtitles for foreign language TV. This entails capturing a program off a TV tuner onto my HD. From there I would export the closed captions and have them robo-translated into English. From there I would want the new text re-imported into the show. That way I could watch the program with these subtitles, which would be full of errors, but still useful. Is this reasonably possible?
posted by Yakuman to Computers & Internet (4 answers total)

My opinion would be that it's possible - I would personally do it under linux, as this is what shell scripting is made for.

So, your problems -

1. Ripping subtitles - Subtitle Ripper does this.

2. Translating the subtitles - this, I'm at a bit of a loss for. Using a web service probably isn't going to be workable, considering the amount of text you'll be translating. OpenLogos looks interesting and useful, but I'm not sure how complete of a product it is. The other problem I see is the issue of formatting - machine translation is sure to introduce some kind of weirdness into the resultant file, and you may run into all kinds of problems when you attempt to re-import it.

3. Re-importing subtitles - Subtitles does this. I'm not sure how well.

4. Platform - Personally, I'd do this in Knoppmyth. It's got good support for HD capture cards, and you can set custom jobs to run after the show is recorded - so your show would record, the script would run, and then the resubtitled show would be imported back into Knoppmyth.

So yeah. I think this is doable, to some extent. It'll depend on how much you can script, and how much patience you have troubleshooting the individual quirks of the software you'll be using.
posted by god hates math at 8:43 AM on November 15, 2008

Another vote here for yes, you could do this. The results will almost certainly be hilariously bad. Go for it!
posted by Hello, Revelers! I am Captain Lavender! at 9:56 AM on November 15, 2008

If you cared to pay for the transalations, you could go out to Amazon's mechanical turk. Use a screengrab and the original text and then have the turkers hand back your translation. This will only work if there are enough people who use mechanical turk who speak your destination language. This will be much higher quality than doing something automatic -- unless this is an art project about the un-translatability of language, your results will be very bad (whether this is a tragedy or a comedy, I leave to you).
posted by zpousman at 12:36 PM on November 15, 2008

YouTube can robotranslate captions if you can get the captions to it.
posted by Ookseer at 3:11 PM on November 15, 2008