How accurate is transcribing with openAI’s #Whisper technology? I tried it on myself but now want to try it on something more substantial. anybody want to share experiences?
We employ Whisper on all of our podcast recordings in Polish. It makes mistakes but so few of them and in a predictable enough way that our listeners are able to follow our episodes without the need to listen to the original recording. I used one of these transcripts to prepare a written article on the matter we were discussing. I use Whisper to transcribe podcasts in foreign languages: French, Swedish, Hungarian, Spanish; I then throw the transcript at DeepL and can read the podcast in a language I can understand so that I can extract the information I need. It works most of the time, it's certainly not a pretty read but it gets me through and I can always extract what I need.
@pvagner Yes, DeepL, my apologies if I did not capitalize that correctly. It's an online translation service powered by neural networks and it makes rather accurate translations even in technical terminology.
Paweł Masarczyk
in reply to Mara Kelland • • •Mara Kelland
in reply to Paweł Masarczyk • • •Peter Vágner
in reply to Mara Kelland • •More about #whisper. I like the fact it can all be run locally.
Paweł Masarczyk
in reply to Peter Vágner • • •Mara Kelland
in reply to Paweł Masarczyk • • •