Follow

My newest video will have subtitles generated by the voice recognition plugin built into . It did a really good job with only a little clean up needed.

· · Web · 5 · 6 · 19

@art I've just begun updating my workflow to do this. I'm sick of waiting for YouTube to generate subtitles just to have me end up needing to add all of the formatting manually. If it's going to be manual, I might as well do it myself!

I'd love to know if there's a way to fix grammar automatically- even title casing the sentences automatically would be a huge help.

@vkc
Would the LanguageTool plugin on Firefox do that since you're editing in a web browser? languagetool.org/
@art

@lps @art I'm not editing in a web browser- I'm using Kdenlive, which is an application installed on the editing machine directly.

When I allow YouTube to do the transcription, it segments it into pieces that aren't grammatically correct most of the time. I've tried online grammar checkers and they do not really streamline this without exporting all of the text and reimporting it (process can take hours on YouTube's side).

@vkc I didnt realize until after I posted it was a response to @art . Ouch, a few hours!? :(

@vkc You can export the SRT file and open in a text editor. I was able to do a lot of cleanup that way. The built-in subtitle editor is pretty slick though. Not sure if that helps.

@art Oooooo this has me scheming.

What if I spun up an offline grammar checker and asked it to ignore the parts of the SRT file that aren't grammar?!?

I remember seeing an open source Python-based grammar engine not too long ago. Someone built it for fixing auto-transcribed lectures. I bet something there could work?

@art OK, now I'm wondering if I could rework something like Vim Grammarous to ignore the SRT-esque parts of the SRT file but still leave them intact for reimportation?

github.com/rhysd/vim-grammarou

@vkc I didn't realize that there is also some basic formatting for the text until I looked it up.

en.wikipedia.org/wiki/SubRip

Not sure how much of that is supported on PeerTube/YouTube.

@art Good!!, didn’t know about this. Tbh, anything that does the synching part must be very helpful, even if the recognition is bad.

@art Oh that's amazing. Any idea what languages #vosk supports?

@esi There are quite a few models to choose from in different languages. I haven't dived much into it. alphacephei.com/vosk/models

@art oh wow that's impressive. a lot more than I expected. Although I've grown used to the disappointment of not seeing my own language in these lists. My language is not very optimal for this kind of stuff given the fact that only about 300-500 thousand people speak it worldwide in total. :P

@art nice. May have to try that again. Last time I tried it I got a total of nothing out of it :/

Sign in to participate in the conversation
Mastodon @ SDF

"I appreciate SDF but it's a general-purpose server and the name doesn't make it obvious that it's about art." - Eugen Rochko