100% offline voice dictation for macOS, Windows & Linux.
Press a shortcut, speak, and your words appear in any app — entirely on your machine.
Website · Download · iOS · Android · Community
- 🔒 100% offline — no network calls, no cloud, no telemetry. Your voice never leaves your machine.
- 🆓 Free & open source — MIT licensed, no subscription, no account required.
- ⚡ Fast & accurate — Whisper (GPU-accelerated) or Parakeet V3 (CPU, ~5× realtime).
- 🖥 Cross-platform — macOS (Apple Silicon & Intel), Windows, Linux.
- ⌨️ Anywhere you type — global shortcut pastes transcription into the focused app.
- Download the latest build from Releases (macOS
.dmg, Windows.msi, Linux.AppImage/.deb). - Launch Dictus Desktop and grant microphone + accessibility permissions.
- Configure your shortcut in Settings (default works out of the box).
- Press, speak, release — your words land in any text field.
Building from source? See BUILD.md.
- Press a configurable shortcut to start/stop recording (or push-to-talk).
- Speak — Silero VAD filters silence in real time.
- Whisper or Parakeet transcribes locally — GPU when available, CPU otherwise.
- Transcribed text is pasted into the focused application.
Dictus runs locally by default. Transcription never leaves your device. Cloud post-processing is OFF by default and opt-in. For the complete list of every outbound network endpoint the app can contact (with how to disable each), see docs/PRIVACY.md.
| Feature | Dictus Desktop | SuperWhisper | Wispr Flow | MacWhisper |
|---|---|---|---|---|
| Price | Free | Free / $8.49/mo | Free / $15/mo | Free / $6.99/mo |
| 100% offline | ✅ | ❌ | ||
| Privacy-first | ✅ | ❌ | ||
| Open source | ✅ | ❌ | ❌ | ❌ |
| Platforms | macOS · Windows · Linux | iOS · macOS · Win | iOS · macOS · Win · Android | iOS · macOS |
| Mobile companion | ✅ (iOS · Android) | macOS-only ecosystem | partial | macOS-only ecosystem |
| Model | CPU | GPU | Notes |
|---|---|---|---|
| Whisper Small / Medium / Turbo / Large | ✓ | ⭐ | GPU recommended for Medium+ |
| Parakeet V3 | ⭐ | — | CPU-optimized, ~5× realtime, auto language detection |
Dictus Desktop is a Tauri app: a React + TypeScript settings UI on top of a Rust core.
whisper-rs— Whisper inferencetranscribe-rs— Parakeet V3 inferencecpal— cross-platform audio capturevad-rs— Silero Voice Activity Detectionrdev/ OS APIs — global shortcuts and paste
- macOS (Apple Silicon + Intel), Windows x64, Linux x64
- Whisper + Parakeet V3 engines
- CLI flags & signal-based control
- Smart Mode Pro — local LLM reformulation
- Custom vocabulary (technical terms, names)
- Audio-file transcription
- Searchable local history
- Sync settings across Dictus iOS / Android / Desktop (offline-first, peer-to-peer)
Open an issue with a feature request — we prioritize the most-upvoted ideas.
Linux — text input tools & global shortcuts
For reliable text input, install the right tool for your display server:
| Display Server | Tool | Install |
|---|---|---|
| X11 | xdotool |
sudo apt install xdotool |
| Wayland | wtype |
sudo apt install wtype |
| Both | dotool |
sudo apt install dotool (add user to input group) |
Runtime library: if startup fails with libgtk-layer-shell.so.0, install your distro's package (libgtk-layer-shell0 on Debian/Ubuntu, gtk-layer-shell on Fedora/Arch).
Wayland shortcuts: must be configured at the desktop-environment level. Use the CLI flags below as the command for your custom binding:
handy --toggle-transcription # toggle recording
handy --toggle-post-process # toggle with post-processing
handy --cancel # cancel current operationSway / i3 example:
bindsym $mod+o exec handy --toggle-transcriptionOr via Unix signals (works with any hotkey daemon):
pkill -USR2 -n handy # toggle transcription
pkill -USR1 -n handy # toggle with post-processingmacOS — CLI usage when installed as a bundle
/Applications/Dictus.app/Contents/MacOS/handy --toggle-transcriptionCLI flags
handy --toggle-transcription # Toggle recording
handy --toggle-post-process # Toggle recording with post-processing
handy --cancel # Cancel current operation
handy --start-hidden # Start without showing the main window
handy --no-tray # Start without the system tray icon
handy --debug # Enable verbose logging
handy --help # Show all flagsThe binary is currently named
handy(a V2 rename is planned).
Whisper (GPU recommended for Medium+):
- macOS — Apple Silicon or Intel
- Windows — Intel / AMD / NVIDIA GPU
- Linux — Intel / AMD / NVIDIA GPU (Ubuntu 22.04 / 24.04 tested)
Parakeet V3 (CPU-only):
- Intel Skylake (6th gen) / AMD equivalent or newer
This project is actively developed. See open issues for current state. Major items:
- Whisper crashes on some Windows / Linux configurations (help wanted — debug logs welcome).
- Wayland support is partial — requires
wtypeordotool(see Linux notes).
Contributions are very welcome. Please:
- Browse open issues —
good first issueandhelp wantedare good starts. - Fork, branch, and submit a PR with a clear description.
- Test on your target platform before requesting review.
- Join the conversation on Telegram or via hello@getdictus.com.
Dictus is free and will stay free. If it saves you time, consider supporting development — it directly funds new features and platform support.
MIT — see LICENSE.
Dictus Desktop is a fork of Handy by cjpais — huge thanks for the foundation.
- OpenAI Whisper — speech recognition model
- whisper.cpp & ggml — cross-platform inference
- Silero VAD — voice activity detection
- Tauri — Rust-based app framework
Made with ❤️ by PIVI Solutions · @getdictus