Embed Notice

I’ve been doing a bunch of interviews over the last year, and using an automated service to get text transcripts from my recordings. It’s been…fine, but the accuracy’s been all over the map.

Recently, @eaton pointed me toward https://github.com/openai/whisper/, and my goodness the accuracy seems *uncannily* good. No speaker identification, sadly — just timestamps — but whisper’s transcripts seem to be such a jump in quality, at least over what I’d been using.