Test frameworks for transcription and diarization of YouTube videos

[finitearth]

Youtube Videos can be a good source of dialogs between interviewers and experts, although the dialog quality of the videos chosen should be considered when choosing which videos could help fine tuning.

Frameworks of consideration could be OpenAIs Whisper for transcription and pyannote for diarization (identification for when speaker switch)

Would be happy to take on this task! :)

[danielpatrickhug]

it would be cool if after transcripts and diarisation same questions can be matched up so that the different responses can be ranked. i.e. Lex's podcast tends to ask guests the same questions to many different people.

[andreaskoepf]

@finitearth ok, this idea was mentioned a few times. I think it is very interesting, please go ahead with your tests and report back .. maybe create an ipynb or some scripts if possible.

[huu4ontocord]

and a readme please :)

[huu4ontocord]

@finitearth wanted to check on status of this issue.

[andreaskoepf]

Closing old data issue.