Meeting Transcription Pipeline

I built a complete meeting processing pipeline that runs entirely locally. No cloud services, no subscriptions, full privacy.

Recording: SoX + BlackHole audio driver captures both system audio and microphone input simultaneously, working with any app (Zoom, Meet, Teams, WhatsApp Web).

Transcription: whisper-cpp provides fast, accurate local transcription with no API costs.

Speaker diarisation: pyannote.audio identifies and labels different speakers in the conversation, producing clean "Speaker A said X, Speaker B said Y" formatted output.

The pipeline is triggered via shell aliases (`rec-start`, `rec-stop`) and processes recordings into structured meeting notes automatically.