Randomly successful AI projects #1
after the release of whisper way back when, voice transcription has been a useful goal, if nothing else for note-taking and reference later. this has been successful in various degrees, but the complication/ideal is to add diarization. whereas projects (and paid programs) have come out, running them on apple silicon successfully has been a challenge.
however with the release of Speech.framework on macOS 26, now there were projects like yap which gave the promise of super-quick voice recognition on mac. but could we add diarization to this?
i set claude on this, and we tried lots of iterations. it seemed to work, but it was super slow, slower than native yap. seems like it wasn’t calling the framework properly. eventually i suggested maybe pivoting to use something like Senko for diarization. it was able to use that instead, and now created a simple and reliable transcription package, and also created all of the documentation and deployment instructions so i could put it on github and have it pretend to look respectable.