WhisperKey lives in your menu bar. Hold a key, speak, release. Your words appear wherever your cursor is. 13 AI models, all running on-device.
Everything runs locally on your machine. No subscriptions, no cloud, no limits.
All speech recognition runs entirely on your machine — Apple Neural Engine, GPU, or CPU. No audio is ever sent to any server.
Choose the right model for your needs: Whisper, Parakeet, Moonshine, SenseVoice, GigaAM. From 31 MB ultra-fast to 1.6 GB highest accuracy.
On supported Macs, use Apple Intelligence to refine your transcriptions — fix grammar, improve clarity, or reformat text automatically.
Text appears wherever your cursor is — Slack, email, Notes, browser, IDE, terminal. Any text field in any app.
Connect OpenAI, Anthropic, or any compatible API to polish transcriptions with custom prompts — summarize, translate, reformat.
Transcribe in English, Chinese, Japanese, Korean, Russian, European languages, and more. Auto-detect or select manually. Translate to English on the fly.
Push-to-talk or toggle mode. Customizable global hotkeys. CLI control via whisperkey --toggle-transcription. Unix signals on Linux/macOS.
Browse past transcriptions, replay audio, and copy previous results. Configurable retention from 3 days to indefinite.
Built-in mic, AirPods, USB headset, external condenser — switch input devices from the menu bar. Clamshell mode support for laptops.
Download once, run forever. Switch models anytime from the settings panel.
Fast and accurate. The recommended default for most users.
Ultra-lightweight. Ideal for quick notes and low-memory machines.
Highest accuracy across all languages. Best for critical transcriptions.
Optimized for Chinese, Japanese, Korean, and Cantonese.
Specialized Russian speech recognition, fast and accurate.
Full range from 58 MB to 1.6 GB. English-only to multilingual. Speed vs accuracy — your choice.
Open the app. Allow Microphone and Accessibility access when prompted. WhisperKey appears in your menu bar and downloads your chosen model.
Position your cursor where you want the text — an email draft, a Slack message, a code comment, anywhere.
The menu bar icon turns purple while recording. Release to stop. Or use toggle mode for hands-free dictation. Customizable hotkey in settings.
Your speech is transcribed on-device in under a second and pasted into the active text field. Optionally refined by Apple Intelligence or an LLM.
6 Model Engines
Whisper, Parakeet, Moonshine, SenseVoice, GigaAM, Apple Intelligence
Tauri + Rust
Native performance with minimal memory footprint. React UI.
100% On-Device
Neural Engine, GPU, or CPU. Zero network calls for transcription.
macOS, Windows, Linux
Apple Silicon, Intel x64, X11, and Wayland support.
17 Languages
English, Chinese, Spanish, French, German, Japanese, Korean, Arabic, and more.
Auto-Update
Built-in updater checks for new versions automatically via GitHub releases.
WhisperKey is a free, open-source voice-to-text app by Evgeny Lisovskiy. Built as part of the Level Up App Factory — an AI-first product studio shipping useful tools powered by on-device intelligence.
Source code on GitHub. Questions or feedback? Reach out at info@levelupbasket.com