User:Ralsettem: Difference between revisions

From SGUTranscripts
Jump to navigation Jump to search
mNo edit summary
No edit summary
Line 1: Line 1:
Self-proclaimed nerd venturing out into the world of transcription supporting a podcast I love to listen to.
Self-proclaimed nerd venturing out into the world of transcription supporting a podcast I love to listen to.


I’m using a local of installation of [https://openai.com/blog/whisper/ Whisper] by OpenAi on [https://github.com/openai/whisper Github] to using the large model to transcribe podcast episodes.
I am using [https://github.com/yinruiqing/pyannote-whisper pyannote-whisper] to create diarization transcriptions. It uses a local of installation of [https://openai.com/blog/whisper/ Whisper] by OpenAi on [https://github.com/openai/whisper Github] to using the large model to transcribe podcast episodes and [https://github.com/pyannote/pyannote-audio Pyannote on Github] to label the different speakers.
It’s theoretically possible to have [https://github.com/lablab-ai/Whisper-transcription_and_diarization-speaker-identification- diarization] of speakers using [https://github.com/pyannote/pyannote-audio Pyannote on Github], however I don’t know python and there is a lack of video tutorials that show the process.
It's about 80% correct most of the time, but there are errors.
Hopefully there will be some kind developers that will create a webUI that is able to transcribe with diarization.

Revision as of 07:34, 7 January 2023

Self-proclaimed nerd venturing out into the world of transcription supporting a podcast I love to listen to.

I am using pyannote-whisper to create diarization transcriptions. It uses a local of installation of Whisper by OpenAi on Github to using the large model to transcribe podcast episodes and Pyannote on Github to label the different speakers. It's about 80% correct most of the time, but there are errors.

Retrieved from "https://www.sgutranscripts.org/w/index.php?title=User:Ralsettem&oldid=17072"