While the source code isn't public for now, you'll find a detailed breakdown of the architecture, decisions, and implementation approach below.
What
A desktop application that records microphone input and transcribes it locally using faster-whisper. Speaker diarization is handled by pyannote.audio. Everything runs on the local machine — no API calls, no cloud.
Why
...
Features
- Local audio recording via sounddevice
- Transcription with faster-whisper (CTranslate2-optimized Whisper)
- Speaker diarization with pyannote.audio
- Desktop GUI built with customtkinter
Stack
| Layer | Choice | |---|---| | Language | Python | | Transcription | faster-whisper | | Diarization | pyannote.audio | | Audio I/O | sounddevice / soundfile | | GUI | customtkinter |
What I learned
...