meeting-transcriber-app

Meeting Transcriber

Native macOS menu bar app that transcribes meetings with speaker attribution — You (microphone) and Them (system audio).

Runs entirely on-device using WhisperKit (CoreML / Neural Engine). No API keys, no cloud, no Python.

Requirements

macOS 14 Sonoma or later
~600 MB free disk space for the Whisper model (downloaded once on first launch)

Installation

Download MeetingTranscriber.zip from the latest release
Unzip and move MeetingTranscriber.app to your Applications folder
Double-click MeetingTranscriber.app to open it
On first launch, the app downloads the Whisper model (~500 MB) — this takes a few minutes depending on your connection

Usage

Click the ⏺ icon in the menu bar:

Action	Description
Start Recording	Begins capturing mic + system audio in 15s chunks
View Transcript	Opens a snapshot of the transcript so far
Set Title…	Names the output file (defaults to “Untitled”)
Stop Recording	Finalises and writes the Markdown transcript
Open Last Transcript	Opens the most recently written transcript
Launch at Login	Toggles auto-start on login
Settings…	Change the output directory

Transcripts are saved as Markdown files (YYYY-MM-DD HH-MM Title.md) in the directory you choose.

Permissions

Microphone — prompted on first Start Recording
Screen Recording — required for system audio (the other side of the call); prompted on first Start Recording

Output format

**[0:01:23] You:** What time works for you?

**[0:01:27] Them:** How about Thursday afternoon?

Building from source

brew install xcodegen
git clone https://github.com/jngo/meeting-transcriber-app
cd meeting-transcriber-app
xcodegen generate
open MeetingTranscriber.xcodeproj

Xcode resolves the WhisperKit SPM package on first open. Build and run with ⌘R.