dee4dog/listen

Listen - offline Windows meeting recorder & transcriber: Whisper transcription (incl. Afrikaans + English translation), speaker separation, recurring-issue tracking, PDF/Word/Excel export, central-server sync # Listen > **Status: pre-alpha — v0.1.0.** Under active testing. Expect rough edges, > and expect breaking changes between releases until 1.0. This app is part of the AIBE Programme. Presented by Akheel Khan (Founder, An Architect, Co-founder Institute of Building Design) Submission by Dirk Cilliers to solve a particular problem using first principle thinking. The app is a pre alpha and still needs major testing. It Is an offline Windows desktop app that records and transcribes speech (English, Afrikaans and 90+ other languages), separates speakers, recognises voices it has heard before, extracts action items and key issues per discussion type, flags recurring issues across sessions, translates speech to English, and exports to PDF, Word or Excel. Everything runs locally on your PC — no cloud account, no API keys, and your recordings never leave your machine (unless you choose to push sessions to your own central server on the local network). ## Features - **Record** from the laptop/desktop microphone, optionally mixed with **system audio** (everything the PC plays) — see the Teams/Meet section below. - **Import** existing audio files (wav, mp3, m4a, mp4, flac, ogg, and more). - **Transcription** with OpenAI Whisper (via `faster-whisper`), model size selectable in Settings (tiny → large-v3). Language auto-detected or fixed. - **Speaker separation**: voices are clustered automatically and labelled Speaker 1, 2, 3… in order of first appearance. - **Name deduction**: self-introductions ("Hi, I'm John…") are used to name speakers automatically where possible. - **Speaker database**: when you name a speaker you can tick *Remember voice*. Their voice profile is stored, and in future recordings they are recognised and named automatically. - **Naming after transcription**: a dialog lets you rename Speaker 1/2/3 to real names (with a ▶ button to hear each voice). Speakers stay editable afterwards: double-click a Speaker cell for a dropdown of the session's speakers (or …