I run a small Dwarf Fortress podcast, and I didn't like the transcription options when we started a few years ago, so I wrote some python glue to do diarization

I run a small Dwarf Fortress podcast, and I didn't like the transcription options when we started a few years ago, so I wrote some python glue to do diarization (separate out speakers) and transcription using a torchaudio project, and either whisper or openai depending on how I'm feeling that day. Works surprisingly well, with timestamps and clean-up: https://github.com/drewbuschhorn/a_strange_mood_podcast_transcripts/blob/main/episode_1/fixup.txt https://github.com/drewbuschhorn/a_strange_mood_podcast_tran... vs https://astrangemoodpodcast.com/2023/01/17/episode-1-is-dwarf-fortress-a-game-or-a-toy/ https://astrangemoodpodcast.com/2023/01/17/episode-1-is-dwar...