Email or username:

Password:

Forgot your password?
PVs- Artist Marcia X's posts Post Back to profile
Top-level
DJ Sundog - from the toot-lab

@CaribenxMarciaX @eloisa @ajroach42 I think that's what Andrew found as well, but this one gets you to 80-90% good quickly and allows you to fine-tune the output captions manually to get it to the quality/accuracy you desire.

11 Nov 2022 at 13:02 | Wall-to-wall | Open on toot-lab.reclaim.technology
3 comments
aschmitz

@djsundog @CaribenxMarciaX @eloisa @ajroach42 Yeah, I've been doing a ton of automatic transcriptions using Whisper (the tool Sundog is talking about) for a variety of videos in the Internet Archive as part of a test run of them. They're not perfect (especially with some names), but they are an excellent starting point, and very fast (and effectively free) to generate. If you want to give it a try, you can throw me a video file link and I'll happily send you some subtitle files.

11 Nov 2022 at 14:49 | Open on ostatus.lardbucket.org
aschmitz

@djsundog @CaribenxMarciaX @eloisa @ajroach42 Oh, right, there was a YouTube video linked in the thread. Here, have some subtitle files: transfer.sh/ELUSuq/Rebelle.srt or transfer.sh/0RIOLO/Rebelle.vtt

You can pick either one of them (I like .srt) and edit it in a text editor to fix up any issues, then you're done! Put it next to the video file with the same name (but different extension) and most players (like VLC) will show the subtitles (you may have to turn on subtitles, depending on the player).

@djsundog @CaribenxMarciaX @eloisa @ajroach42 Oh, right, there was a YouTube video linked in the thread. Here, have some subtitle files: transfer.sh/ELUSuq/Rebelle.srt or transfer.sh/0RIOLO/Rebelle.vtt

You can pick either one of them (I like .srt) and edit it in a text editor to fix up any issues, then you're done! Put it next to the video file with the same name (but different extension) and most players (like VLC) will show the subtitles (you may have to turn on subtitles, depending...

11 Nov 2022 at 15:15 | Open on ostatus.lardbucket.org
Andrew (bookseller era)

@djsundog @CaribenxMarciaX @eloisa

We're using Whisper as a base for generating subtitles, and then cleaning them up by hand in subtitle composer.

@Ethancdavenport is doing the majority of the manual work in cleanup.

Whisper's auto gen is pretty good, it's not perfect, but it takes generating subtitles from a multi-day thing down to a few hours for us.

I'll also boost it up in case anyone has better ideas.

11 Nov 2022 at 16:24 | Open on retro.social
Go Up