Some transcription service will spit out unicode zero width space characters, which looks like
<200b><200c><200b><200c><200b><200c><200b><200c><200b><200c><200b><200c><200b><200c><200b><200c><200b><200c><200b><200c><200b><200c>
The following sed command will fix it:
gsed -i 's/\xe2\x80\x8b//g' *.srt #<200b>
gsed -i 's/\xe2\x80\x8c//g' *.srt #<200c>