In one line
Fully local pipeline that turns lecture recordings into transcripts, glossaries and summaries overnight.
Context & problem
Long lecture recordings – I needed summary, searchable notes and a glossary without re-watching hours of audio.
What I did
- Capture on the laptop (BlackHole + ffmpeg), processing on the always-on Mac Mini
- whisper.cpp (large-v3) for transcription, llama.cpp for summaries and glossary
- Output into OneNote
- Solved large recording files and Whisper repetition loops
Result / impact
2 hours saved per lecture. Looking up notes and concepts became much faster.
Details
- Compliance & data protection: lecturers' and classmates' voices never leave my own hardware; no cloud costs