I have a handful of audio and video recordings of my parents and other relatives for which I would love to have transcriptions. In the spring I found a (not free) web site called
https://veed.io to which you can upload audio or video files and it will attempt to auto-generate subtitles. The resulting quality varies, mostly between decent and shockingly good. I have only tried detecting English speech. (I have one recording where the sound quality is poor and that one totally failed to produce useful results.)
Rough steps after creating an account:
- Click New Project
- Upload your audio/video
- In the left panel, click Subtitles
- Click Auto Subtitles, select the language, and click Start
- Wait a while (it seems to take a few minutes for every half-hour of source material)
- The subtitles will eventually appear in the Subtitles panel
- Click Options
- Click Download .txt
- Now you will download the a text file containing subtitles
So far I'm finding I need to spend between an hour and maybe four hours per hour of original recording doing cleanups, fixing transcription errors, formatting the result, etc.
Quality of the translation will vary based on a lot of factors, mostly audio recording quality, whether it's more of a narration than a multi-person overlappping conversation, how strong the speaker's accent is, etc. Naturally, foreign place names and proper names are more likely to be mangled.
Still, this is a great way to kickstart a quarantine-project. Based on some past transcribing I've done, this cuts the time/effort needed down to about a quarter of what would be needed without this, give or take.
Veed.io is $30/month and that gives you 4 hours of automatic subtitle generation. If you run out, you can top up the hours ($10 for 5 hours) or wait for the next month. (The top-up button is currently broken, but you can send them an email.)
(I found this web service in the spring helping some college students who needed transcriptions of some Zoom lectures.)