Why I stopped uploading my meetings to the cloud and built my own transcriber

Описание и применение

Автор создал Fennec, чтобы хранить аудио и транскрипты локально, интегрировать их с кодинг‑агентами и избежать неопределённости и расходов облачных сервисов.

Задача

Uncertainty about where raw meeting audio is stored, how long it stays, and whether it is used for training or retained after deletion, plus difficulty accessing transcripts for coding agents and high cost of cloud services.

Результат

Audio stays on the local machine, transcripts and minutes are plain files, workflow is fully local except for optional cloud LLM calls, eliminating data‑privacy concerns and reducing recurring costs.

Рабочий процесс

• 1. Record meeting using Fennec’s system audio tap and mic input.
• 2. Transcribe audio with whisper.cpp (or Apple on‑device ASR on macOS).
• 3. Perform speaker diarization with sherpa‑onnx.
• 4. Optionally translate live output on macOS.
• 5. Store all outputs (audio, transcript, minutes.md, metadata) in a dedicated session folder.
• 6. When ready, press "Create Minutes" to send transcript text to a chosen coding agent CLI (Claude Code, Codex, Gemini CLI).
• 7. Receive summarized minutes and save locally; raw audio remains on device.
• 8. Backup session folders with any file‑based backup tool; search with grep or editor.
• 9. Optionally rebuild Fennec in Rust/Tauri for cross‑platform use.

Доказательства

Recording, transcription, speaker diarization and translation all run on the local machine.

Источник

The raw audio never leaves. The transcript leaves only when you press the button, and only to the agent you set up.

Источник