Local dictation for macOS
Speak. VoiceLah types in the app you are using.
Start recording with a global shortcut, speak, and get your text wherever the cursor is. Core transcription runs directly on your Mac.
Free download · macOS 14 or later · Apple Silicon and Intel
Speak naturally. VoiceLah inserts the text into the app you were using.
How to use it
1. Place the cursor. Open Mail, Notes, your browser, or any other app where you want to write.
2. Start and speak. Use your configured shortcut. The indicator and visual effects show when VoiceLah is recording.
3. Stop recording. VoiceLah transcribes the audio on your Mac and inserts the result wherever you were typing.
What it does
Fast dictation without switching apps.
It records, transcribes locally, and returns the text to the application that was active before dictation.
Global shortcut
Start and stop recording without leaving Mail, Notes, an editor, or your browser.
Visual feedback
A recording pill, waveform, and audio-reactive effects clearly show when the microphone is listening.
Local Whisper
Choose a model that balances speed, memory use, and accuracy directly on your device.
Text where you need it
VoiceLah remembers the active app and inserts the transcription wherever you were working.
Optional AI
Grammar refinement is separate from transcription and can use an API key supplied by the user.
Native Mac app
SwiftUI and AppKit work together for the menu bar, floating windows, permissions, and system integration.
Under the hood
From your voice to the cursor, without unnecessary steps.
Settings
Configure recording, quality, effects, and AI.
The main options are organized under the General, Effects, and AI tabs.
Shortcut and recording mode
Choose a global key combination and use Hold to Record, or press once to start and again to stop.
Transcription quality
Select Fast, Balanced, Accurate, or Best to balance speed, accuracy, memory use, and model size.
Recording effects
Enable or disable the effect, then choose its style, screen position, color, size, and opacity.
Startup and sounds
Choose whether VoiceLah opens at login and whether it plays sounds when recording starts and stops.
Optional AI refinement
Choose OpenAI, OpenRouter, Anthropic, or local Ollama, then configure the model, style, and custom prompts.
Permissions and onboarding
Check Accessibility permission, open the macOS system prompt, and restart onboarding whenever needed.
The part you do not see
VoiceLah worked. But the way it inserted text conflicted with Mac App Store rules.
To write in the active window, this kind of app must interact with Accessibility permissions and, as a fallback, macOS automation. These are powerful capabilities that must be clearly explained and explicitly requested.
The version prepared for the Store was rejected. Instead of hiding the problem, I recovered the working build, rebuilt the permission onboarding, and removed the sandbox from the new version. Direct distribution preserves the core feature, but it still requires Developer ID signing, notarization, and a serious release process.
It was the most useful lesson from the project: on macOS, the product is not just the interface. It is also entitlements, TCC, code signing, window focus, automation, and user trust.
Privacy
Local transcription. AI only when you enable it.
The core feature requires neither an account nor sending audio to a remote service.
Transcription on your Mac
WhisperKit processes the recording on the device and produces the text locally.
- No account required
- Transcription history stored locally
- No audio uploaded
- Selectable Whisper model
Optional AI refinement
If enabled, your text can be corrected or rewritten by the provider you configure.
- OpenAI, OpenRouter, or Anthropic
- Ollama for local AI processing
- Custom styles and prompts
- API keys stored in the macOS Keychain