Whisperheim is the audio Swiss army knife for Windows 11. Local dictation, call transcription with speaker separation, voice-message transcription. No cloud, no subscription, no internet at runtime.
Grab the latest WhisperHeim-*-win-Setup.exe from the
Releases page.
The Release notes always include the SHA-256 of Setup.exe so you can verify
the download (certutil -hashfile WhisperHeim-*-win-Setup.exe SHA256 in any
terminal).
Whisperheim is currently unsigned (see why), so Windows will warn you on first run. Two paths:
- Windows Defender SmartScreen — the dialog reads "Windows protected your PC". Click More info, then Run anyway.
- Smart App Control — if the dialog reads "Smart App Control" instead of "Windows Defender SmartScreen", there is no override. SAC is opt-in but on by default on some fresh Windows 11 25H2 installs. You'll need to either turn SAC off system-wide (irreversible until OS reinstall) or wait for a signed build. Signing is on the roadmap post-UG registration — see docs/why-unsigned.md.
A short screen recording of the click-through lives in
docs/media/install.mp4 once recorded.
- A small setup window opens and downloads the ~640 MB Parakeet ASR model
from Hugging Face. One-time download; cached in
%APPDATA%\WhisperHeim\models. You can skip and configure later — dictation will then prompt for the download on first use. - Whisperheim drops to the system tray. Right-click the tray icon for the menu and settings.
- The Silero VAD and Pyannote segmentation models ship bundled inside the installer; you don't have to download those.
| Hotkey | What it does |
|---|---|
Ctrl + Win |
Hold to dictate into the focused text field |
Ctrl + Win + R |
Toggle call recording (system audio + microphone, with speaker separation) |
Ctrl + Win + Alt |
Trigger a template by speaking its name |
All hotkeys are remappable from the General settings page.
For YouTube and audio-stream transcription (and as a fallback OGG/Opus
decoder), Whisperheim needs FFmpeg on PATH. Whisperheim does not bundle
FFmpeg — the user picks the build and license. The easiest path:
winget install -e --id Gyan.FFmpegThe first time you invoke a feature that needs FFmpeg, Whisperheim shows a modal with the same command and an "Open download page" link. Drag-and-drop transcription of common WAV/MP3 files keeps working without FFmpeg.
By default, everything user-generated lives under:
%APPDATA%\WhisperHeim\
settings.json,bootstrap.json— your settingsrecordings\— raw WAV capturestranscripts\— transcribed textmodels\— the downloaded Parakeet model
You can move the data root to any folder you like (e.g. a synced OneDrive/Dropbox/Google Drive folder) from the General settings page.
Uninstall does NOT delete this data. If you want a fully clean removal,
delete %APPDATA%\WhisperHeim\ and any custom data folder yourself after
uninstalling. Whisperheim drops a WhisperHeim-data-location.txt on your
desktop on uninstall to remind you where things live.
- Windows 11 (24H2 or later recommended), x64
- ~2 GB RAM idle, ~3 GB during transcription
- ~700 MB free disk for the Parakeet model
- A working microphone (and a loopback-capable audio device for call transcription — any standard Windows 11 setup qualifies)
- All transcription runs locally on your machine. No audio, transcripts, or metadata leave the device at runtime.
- No telemetry. Whisperheim does not call home — no analytics, no crash reporting, no usage pings.
- No accounts. No login, no cloud, no API keys required for the core features.
- The only outbound traffic on a fresh install is the one-time Parakeet model download from the Hugging Face CDN. Auto-updates check GitHub Releases.
- Optional integrations you opt into (Ollama for AI analysis, a synced data folder via your existing cloud client) are clearly labelled as such.
Whisperheim is not yet code-signed. Microsoft Trusted Signing isn't available to individual developers in Germany at the time of writing; signing is planned post-UG (Unternehmergesellschaft) registration. Until then:
- Every Release publishes the SHA-256 of
Setup.exein the Release notes. - The full source is in this repository —
git logis your audit trail. - See docs/why-unsigned.md for the full story on SmartScreen, Smart App Control, and what the signing roadmap looks like.
git clone https://github.com/heimeshoff/WhisperHeim.git
cd WhisperHeim
dotnet build src/WhisperHeim/WhisperHeim.csproj
dotnet run --project src/WhisperHeim/WhisperHeim.csprojRequires the .NET 9 SDK.
For the release pipeline (Velopack pack, GitHub Actions, SHA-256 surfacing),
see docs/release.md.
If Whisperheim is useful to you and you want to chip in, there's a tip jar:
TBD — the licensing decision is pending the monetization research, see the roadmap. The repository currently carries an Unlicense declaration; this may change before the first public 1.0 release.