Hotkey-triggered speech-to-text for Windows. Tap the hotkey to start dictating, tap again to stop — the transcription is auto-pasted into the focused window.
Inspired by whisper-writer, but built native for Windows for two reasons:
- A UI with a bit more personality.
- Windows' built-in dictation handles Korean poorly — especially when Korean speech is mixed with English loanwords (which is most real-world Korean).
- Most existing tools can't use the Windows key as part of the activation hotkey. WinWhisper can.
Powered by OpenAI Whisper via Whisper.net (local) or the OpenAI API (cloud).
Requires Windows 10 build 17763 (1809) or later, x64, and the .NET 9 SDK.
git clone https://github.com/seheepeak/WinWhisper.git
cd WinWhisper
powershell -ExecutionPolicy Bypass -File install.ps1install.ps1 publishes a Release build directly into %LocalAppData%\Programs\WinWhisper and drops a shortcut in the Startup folder so it auto-starts on next login. Re-run any time to update.
- Press the hotkey (default
Win+~) to start recording, then speak. - Press it again to stop. The transcription is auto-pasted into the focused window.
Escwhile recording to cancel.- Right-click the tray icon for Settings, "Copy last transcription", or Exit.
Other recording modes (hold-to-record, voice-activity, continuous) are available in settings.json under recording.recordingMode.
First launch opens Settings — pick a local Whisper model to download or paste an OpenAI API key.
To change the hotkey, edit recording.activationKey in %LocalAppData%\WinWhisper\settings.json and restart. Examples: "win+~", "ctrl+shift+v", "alt+space".
See CLAUDE.md.
MIT.