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 now supports multiple transcription providers and lets you choose the model you want per provider:
- Groq
- Fireworks AI
- ElevenLabs
- Deepgram
- AssemblyAI
If you want the easiest free setup, Groq is still a great default. Its free tier requires no credit card, and Mumble can also use a saved Groq key for optional Smart Formatting.
- 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.
- Provider and model choice — pick the transcription provider and model that fit your workflow.
- 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 optional Groq-backed 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 at least one supported transcription provider API key. If you want Smart Formatting, save a Groq key as well.
The app will walk you through the rest.
For contributors or if you prefer to build locally — you'll need Xcode 15+ and at least one supported transcription provider 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 Selected provider Optional formatting Text appears
┌──────────┐ ┌─────────────┐ ┌────────────────┐ ┌────────────────┐ ┌───────────┐
│ Hold Fn │ ──▶ │ Record mic │ ──▶ │ STT API + model│ ──▶│ Groq or local │ ──▶│ At cursor │
└──────────┘ └─────────────┘ └────────────────┘ │ context format │ └───────────┘
└────────────────┘
▲
┌──────┴───────┐
│ Detect app: │
│ email? chat? │
│ code? other? │
└──────────────┘
Under the hood: DictationManager orchestrates everything — shortcut monitoring, audio recording, provider-specific transcription, optional Groq-backed Smart 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.