Talk to your Mac. It types back.
A tiny, completely free menu bar app that turns your voice into text — instantly.
Hold Fn → speak → release → text appears at your cursor.
Exactly the way you wanted it.
Signed & notarized · macOS 14+
Because typing is overrated. Mumble sits quietly in your menu bar and lets you dictate into any text field on macOS. It figures out where you are — Slack, email, iMessage, a Google Doc — and adjusts the tone to match.
Casual message to a friend? Mumble keeps it chill. Work email? Mumble cleans it up. You just talk.
Mumble is powered by Groq's whisper-large-v3 model — the most accurate open speech-to-text model available, running at lightning speed on Groq. Groq's free tier requires no credit card and gives you:
- 2,000 requests/day — that's 2,000 individual dictations
- 8 hours of audio/day (28,800 seconds) — more than enough for all-day use
- 20 requests/minute — plenty for natural dictation
Just grab a free API key from console.groq.com and you're set. No subscriptions, no trials, no catch.
- Push-to-talk with Fn — hold to record, release to transcribe. No clicking, no waiting.
- Instant text insertion — transcribed text drops right into whatever app you're using.
- 7-day recovery history — every completed transcription is kept locally for 7 days in a dedicated History tab so you can recover text if paste misses.
- Context-aware tone — Mumble detects the active app and adjusts formatting automatically.
- Blazing fast transcription — powered by Groq's
whisper-large-v3for maximum accuracy at incredible speed. - Zero clutter — lives in the menu bar, no Dock icon, no windows in your way.
- Guided setup — onboarding walks you through API key, permissions, and preferences.
- Custom dictionary — add vocabulary corrections so Mumble always spells your names, jargon, and brand terms the right way.
- Smart formatting — an LLM pass automatically structures your dictation based on the active app (email gets greeting/body/sign-off, messaging stays casual with emoji, code preserves technical terms, and everything else gets clean punctuation). Detected from the frontmost app's bundle ID and page title; toggleable in Settings.
Grab the latest release from GitHub Releases — the app is signed and notarized, so you can run it straight away. Requires macOS 14+ (Sonoma) and a free Groq API key (no credit card required).
The app will walk you through the rest.
For contributors or if you prefer to build locally — you'll need Xcode 15+ and a free Groq API key.
brew install xcodegen # one-time setup
cd /path/to/Mumble
xcodegen generate # generate the Xcode project
open Mumble.xcodeproj # Cmd+R to build and run You speak Mumble listens Groq transcribes Smart formatting Text appears
┌──────────┐ ┌─────────────┐ ┌────────────────┐ ┌───────────────┐ ┌───────────┐
│ Hold Fn │ ──▶ │ Record mic │ ──▶ │ Whisper API │ ──▶│ Context-aware │ ──▶│ At cursor │
└──────────┘ └─────────────┘ └────────────────┘ │ LLM format │ └───────────┘
└───────────────┘
▲
┌──────┴───────┐
│ Detect app: │
│ email? chat? │
│ code? other? │
└──────────────┘
Under the hood: DictationManager orchestrates everything — shortcut monitoring, audio recording, Groq API transcription, context-aware formatting, and text insertion via Accessibility APIs.
| Problem | Fix |
|---|---|
| Fn key not working | System Settings > Keyboard > set "Press fn key to" → Do Nothing |
| No accessibility permission | System Settings > Privacy & Security > Accessibility > enable Mumble |
| No microphone permission | System Settings > Privacy & Security > Microphone > enable Mumble |
| Text not inserting | Make sure a text field is focused in the target app before recording. If a paste still misses, open History in Settings and copy the saved transcription. |
Rebuilt or moved the app? You may need to remove the old Accessibility entry and re-add it.
- Fn key can conflict with certain macOS keyboard settings
- Browser URL detection doesn't work with all browsers
- Requires an internet connection (no offline fallback yet)
- Only Fn key is supported for now (custom hotkeys coming soon)
- Custom hotkey support
- Multi-language transcription
- Local Whisper model for offline use
- User-defined tone profiles
See LICENSE for details.