when people just paste their script into the video and let the automated systems match up the script to when things are said
but those don't always match up, so you have things added not in the script that therefore aren't subtitled, so it stays on one line that was said 30 seconds ago for awhile
and other times, parts of the script are cut, and the automated systems have to skip through several lines of text at once to get to the relevant bit
and yet-other times one line is replaced with a similar but very different line and you have to discern yourself what info to pay attention to between the two streams of dialogue