Real Time Sub-Titles For Videos : Thoughts, Ideas & Algorithms?

This novel thought just struck my mind and I'm sure it's going to be a 'HUGE' thing if implemented. The idea is like this :-

When a video streams, the algorithm will quickly fetch the streamed data; extract the audio signals out of it and depending upon the voice, add sub-titles in the language of user's choice. So as soon as you start streaming your video; the sub-titles will start appearing on user's screen in the language decided by the user.

What do you think? Quite awesome? Already done? How to do it? Algorithm tryouts? Bring them on! 😁

Aim: ^^^
Apparatus: Super fast speech recognition system which needs self leaning computing system + Google translate.
Major user: frequent visitor of YouTube video learner and PirateBay. ☕


Edit: here is one of this kind project developed but I din't find how actually it's working
https://blog.mathieu-leplatre.info/announcing-subtivals-realtime-subtitles-for-film-festivals.html

^^ That is still not completely real-time. The subtitles should still be prepared beforehand. It's just projected from a different projector, so that even if the original video is not subtitled, we can still have our own subtitles superimposed from another projector. And then there are other features like sncing, speeding up the subtitles etc...

Truly real-time subtitle generation would be quite a task, I think. Speech recognition, Natural Language processing and translation. All the tasks are formidable in themselves. Should be combined in real time. But the idea is really good.