4 ms·
Just using Whisperv3 high-quality to dump all the transcripts and make them mostly searchable would be a big improvement over the raw audio.
by gwern 3y ago
Just using Whisperv3 high-quality to dump all the transcripts and make them mostly searchable would be a big improvement over the raw audio.
- nerdponx 3y agoAnd once you have relatively clean text, you could do fun things like topic modeling to segment "interview content" discussing the work/career/etc of the interviewee, from "banter".