HushScribe — Local Meeting Transcription for macOS

Privacy first

Nothing leaves your Mac.

By design, not by policy. Every component runs locally. There are no servers to breach, no accounts to leak.

Audio never touches a server Transcription runs on-device via Apple Silicon's Neural Engine. Your voice stays on your computer, always — not in transit, not cached remotely.
AI that stays local Summaries use Qwen3, Gemma 3, or Apple's NaturalLanguage framework — all running on-device. No API key, no external request, no account.
You own the output Transcripts are plain .md files in a folder you choose. No proprietary format, no sync service, no vendor lock-in.
Zero telemetry No analytics, no crash reporting, no usage tracking. The app makes zero network requests while running. Verified open source.
Hidden from screen sharing The app window is invisible to screen recording by default. It won't appear on your call, in screenshots, or in recordings taken by others.
Open source Every line is on GitHub. Audit the code, build it yourself, or fork it. No black boxes — privacy claims are verifiable, not a marketing promise.

Features

Multilingual transcription Parakeet-TDT v3 — 25 European languages, auto-detected, runs on Apple Silicon's Neural Engine. Fastest transcription option.
Multiple transcription models Parakeet (default), WhisperKit Base, WhisperKit Large v3, or Apple Speech. Switch at any time. All on-device.
Call Capture Captures mic and system audio simultaneously. Detects Teams, Zoom, Slack and filters audio to just that source.
On-device AI summary Open any transcript and click Generate Summary. Qwen3, Gemma 3, or Apple NL — all local, no API key. Supports custom prompts.
Speaker diarization Splits remote audio into labelled speakers after the session ends. Assign real names before saving to your vault.
Auto-record meetings Detects active calls via app presence and mic activity. Starts and stops recording automatically — no button to forget.
Obsidian vault native output Structured .md with YAML frontmatter. Drop it directly into Obsidian or any folder-based workflow.
Export transcripts Export any session as .srt subtitles or structured .json directly from the Transcript Viewer.
Lives in the menu bar No dock icon. Runs quietly in the background, accessible from the status bar at any time.
Self-contained No Ollama, no local server, no background daemons. Everything runs inside the app — just download and go.
Voice Memo mode Mic-only recording saved to a separate folder. Voice notes never mix with meeting transcripts.

Output

Every recording produces a structured Markdown file. Additionally you can export to other formats.

You can summarise the transcript or export to other formats.

.srt — Subtitle export .json — Structured transcript

On-device AI summary

Obsidian-ready markdown file

---
date:         2026-04-25
duration:     47m
participants: [You, Sarah, Marcus]
type:         call-capture
---

## Summary
Discussion covered Q2 roadmap priorities
and the migration timeline for the auth
service. Action items assigned to Marcus.

## Transcript
[00:01] You: Let's start with the roadmap...
[00:14] Sarah: The auth migration is blocking us.
[00:31] Marcus: I can take that next sprint.
[01:04] You: Great — let's lock that in.

Install

Requires macOS 26+ and Apple Silicon. No account, no license key, no recurring fees.

Homebrew (recommended)

brew tap drcursor/hushscribe https://github.com/drcursor/HushScribe
brew install --cask hushscribe

To update: brew upgrade --cask hushscribe

Manual install

Download the DMG, open it, drag HushScribe to /Applications.

Download DMG

Legal notice: Laws regarding recording conversations vary by jurisdiction. In many places, all parties must consent before a conversation may be recorded. It is your sole responsibility to ensure compliance with applicable local laws. The authors of HushScribe accept no liability for unlawful use.

Your meetings. Your data.

Nothing leaves your Mac.

Your meetings.
Your data.