Hacker News new | ask | show | jobs
by dale-cooper 2992 days ago
I experimented with something related, but instead of fixing subtitles i focused on filtering out bad ones. My approach was:

* Extract a few sentences from the middle of the subtitle file.

* Extract the audio for those sentences from the video file

* Run speech-to-text and verify that at least a few of the words were the same.

It kind of worked on some shows, but not reliably..