Local speech-to-text for Windows using OpenAI Whisper — like Wispr Flow, but free and fully offline.
- Push-to-Talk & Toggle Mode — customizable global hotkey (default:
Win+D) - 100% Local — no cloud, no API keys, complete privacy
- Single Process — one .NET executable, no Python, no FFmpeg, no Node.js
- GPU Acceleration — CUDA support for NVIDIA GPUs with automatic CPU fallback
- Auto-Paste — transcribed text is copied to clipboard and pasted at cursor
- Multiple Languages — German, English, and many more, or auto-detect
- System Tray — runs quietly in the background
- Floating Status Bar — always-visible recording indicator
- Auto-Updates — delta updates via Velopack
- Windows 10/11 (64-bit)
- NVIDIA GPU with CUDA (optional, recommended for faster transcription)
Download the latest installer from Releases.
The installer handles everything — no dependencies to install manually.
git clone https://github.com/StaniB88/WhisperVoice.git
cd WhisperVoice
dotnet run --project src/WhisperVoicedotnet publish src/WhisperVoice -c Release -r win-x64 --self-contained -o publish- Launch — the app opens with a setup wizard on first run
- Hold Hotkey — press and hold your configured hotkey (default:
Win+D) - Speak — your audio is recorded locally
- Release — audio is transcribed using Whisper
- Done — text is pasted at your cursor position
Settings are stored in %APPDATA%/whisper-voice/config.json
| Option | Description |
|---|---|
| Whisper Model | tiny (fast) → large (best quality) |
| Language | German, English, auto-detect, and more |
| Hotkey | Any key combo, recorded in settings |
| Toggle Mode | Press once to start, press again to stop |
| Auto-Paste | Automatically paste with Ctrl+V after transcription |
| Floating Bar | Show/hide the floating status indicator |
| Theme | Multiple dark themes to choose from |
Models are downloaded on first use and stored in %APPDATA%/whisper-voice/models/.
| Model | Size | VRAM | Speed | Quality |
|---|---|---|---|---|
| Tiny | ~75 MB | ~1 GB | ~10x | Basic |
| Base | ~142 MB | ~1 GB | ~7x | Good |
| Small | ~466 MB | ~2 GB | ~4x | Better |
| Medium | ~1.5 GB | ~5 GB | ~2x | Great |
| Large v3 | ~2.9 GB | ~10 GB | 1x | Best |
| Large Q5_0 | ~1.1 GB | ~5 GB | ~1.5x | Great |
| Large Turbo | ~1.5 GB | ~6 GB | ~8x | Best |
| Large Turbo Q5_0 | ~574 MB | ~3 GB | ~10x | Great |
English-only variants (Tiny, Base, Small, Medium) are also available for improved accuracy on English speech.
Single-process .NET 10 application — no child processes, no IPC, no HTTP servers.
User presses hotkey
│
▼
Win32 SetWindowsHookEx (global keyboard hook)
│
▼
NAudio (16kHz/16-bit/mono WAV capture)
│
▼
Whisper.net (CUDA → CPU fallback)
│
▼
Clipboard + SendInput (auto-paste at cursor)
| Component | Technology |
|---|---|
| UI | Avalonia UI 12 |
| MVVM | CommunityToolkit.Mvvm |
| Transcription | Whisper.net (whisper.cpp bindings) |
| Audio | NAudio |
| Hotkeys | Win32 P/Invoke |
| Updates | Velopack |
- Use a smaller model (Tiny or Base)
- Install NVIDIA CUDA drivers for GPU acceleration
- Ensure GPU drivers are up to date
- Some key combinations are reserved by Windows
- Try a different hotkey in Settings → Hotkey → Record
- Check your internet connection
- Models are downloaded from Hugging Face — ensure it's not blocked
Copyright (C) 2026 AnyAutomation.
Whisper Voice is licensed under the GNU Lesser General Public License v3.0 or later (LGPL-3.0-or-later). See LICENSE / COPYING.LESSER and COPYING for the full license texts.
Third-party components are listed in THIRD-PARTY-NOTICES.md.
- Whisper.net — C# bindings for whisper.cpp
- Avalonia UI — cross-platform UI framework
- NAudio — .NET audio library
- Velopack — installer and auto-update framework
- OpenAI Whisper — speech recognition model
If you find Whisper Voice useful, consider buying me a coffee.