What is this?
quill is a minimal, fully local macOS meeting recorder and transcriber. One menu-bar click records your mic and all system audio; when you stop, it transcribes both tracks on-device into a speaker-tagged transcript.
quill is a minimal, fully local macOS meeting recorder and transcriber. One menu-bar click records your mic and all system audio; when you stop, it transcribes both tracks on-device into a speaker-tagged transcript.Is this a menu-bar app or a command-line tool?A single Swift binary with no app bundle, whose entire UI is the menu-bar tray icon. You launch it from a terminal or via a LaunchAgent, then drive it by clicking the feather.
A single Swift binary with no app bundle, whose entire UI is the menu-bar tray icon. You launch it from a terminal or via a LaunchAgent, then drive it by clicking the feather.How do I record my first meeting?Run quill in a terminal or via the LaunchAgent, click the feather and choose Start recording — first use prompts for microphone and System Audio Recording permissions. Stop when the meeting ends; transcription starts automatically and a notification fires when the transcript is ready.
Run quill in a terminal or via the LaunchAgent, click the feather and choose Start recording — first use prompts for microphone and System Audio Recording permissions. Stop when the meeting ends; transcription starts automatically and a notification fires when the transcript is ready.What does my Mac need to run it?macOS 15 or newer, since system audio is captured with Core Audio process taps — no virtual device or kernel extension. Apple Silicon is recommended, and ~600 MB of transcription models download once on first use; quill doctor tells you whether they are already cached.
macOS 15 or newer, since system audio is captured with Core Audio process taps — no virtual device or kernel extension. Apple Silicon is recommended, and ~600 MB of transcription models download once on first use; quill doctor tells you whether they are already cached.Any gotchas before I use it in a real meeting?The system-audio tap is global: it records everything your Mac plays, so don't run Spotify or other sound sources during a meeting. Silent recordings usually mean the Screen & System Audio Recording permission is off.
The system-audio tap is global: it records everything your Mac plays, so don't run Spotify or other sound sources during a meeting. Silent recordings usually mean the Screen & System Audio Recording permission is off.Does my audio ever leave my machine?No — recording, storage and transcription all happen on-device. The only download is the one-time fetch of the transcription models.
No — recording, storage and transcription all happen on-device. The only download is the one-time fetch of the transcription models.What files does one session leave behind?Each session lands in ~/Recordings: mic.caf, system.caf, meta.json, transcript.json (timed, speaker-tagged segments), transcript.md for reading and transcribe.log. Audio is CAF, which needs no finalization pass — if the process dies mid-meeting, everything already written stays readable.
Each session lands in ~/Recordings: mic.caf, system.caf, meta.json, transcript.json (timed, speaker-tagged segments), transcript.md for reading and transcribe.log. Audio is CAF, which needs no finalization pass — if the process dies mid-meeting, everything already written stays readable.Why record the mic and system audio as two separate tracks?Speech models do better on clean single-source audio, and mic-vs-system is free two-party diarization — 'me vs them' with no speaker-identification model. Each track is transcribed separately and merged onto one clock by timestamp.
Speech models do better on clean single-source audio, and mic-vs-system is free two-party diarization — 'me vs them' with no speaker-identification model. Each track is transcribed separately and merged onto one clock by timestamp.How fast is transcription, and can I keep recording while it works?Roughly 20 seconds per hour of audio on Apple Silicon with the default Parakeet engine. Jobs run in a serial queue, so you can start a new recording while the last one still transcribes, and unfinished jobs resume on the next launch.
Roughly 20 seconds per hour of audio on Apple Silicon with the default Parakeet engine. Jobs run in a serial queue, so you can start a new recording while the last one still transcribes, and unfinished jobs resume on the next launch.Does it handle languages other than English?Not yet — the built-in Parakeet engine is English-only. A WhisperKit large-v3-turbo engine is planned as a fallback and re-transcription option, and other languages are expected to come with it.
Not yet — the built-in Parakeet engine is English-only. A WhisperKit large-v3-turbo engine is planned as a fallback and re-transcription option, and other languages are expected to come with it.What are similar projects?DigUp, parrot, better-voice, lipflow, meetily, call.md, murmur-youtube, rooms
DigUp, parrot, better-voice, lipflow, meetily, call.md, murmur-youtube, roomsWhich topics is it in?meeting-notes, speaker-diarization, recording, menubar-app, on-device, transcription, speech-to-text, native, macos-app, apple-silicon, audio, swift
meeting-notes, speaker-diarization, recording, menubar-app, on-device, transcription, speech-to-text, native, macos-app, apple-silicon, audio, swift