Voice Typeless
Voice-typeless (VTL)
Say it. Type less. — Zero-dependency, fully offline voice-to-text for every app.
Voice-typeless (VTL) is a lightweight desktop app that captures your voice with a global hotkey and instantly types the transcribed text into any focused application — no internet connection, no account, no cloud.
📸 Screenshot placeholder — UI screenshot coming in v0.2.0 release
Why This Exists
Dictation tools either require a cloud connection (breaking privacy) or are locked to specific apps. VTL runs 100% offline using on-device speech models (SenseVoice / Whisper via sherpa-onnx), works in any text field on Windows, and restores your clipboard after pasting so you never notice it was used.
Quick Start
Prerequisites
| Requirement | Version | Notes |
|---|---|---|
| Rust + Cargo | 1.77+ | MSVC toolchain on Windows |
| Node.js | 18+ | |
| Go | 1.23+ | |
| VS 2019 Build Tools | MSVC 14.29 | See note below |
Windows toolchain note: Use VS 2019 Build Tools (MSVC 14.29). MSVC 14.50 (VS 2025) is missing
lib\x64\andinclude\directories required by the Rust linker. Run. .\scripts\env-msvc.ps1to auto-detect and apply the correct toolchain environment.
Build & Run
# 1. Activate the correct MSVC environment (Windows only)
. .\scripts\env-msvc.ps1
# 2. Install frontend dependencies
cd frontend && npm install && cd ..
# 3. Start development mode (hot-reload Tauri + Svelte)
cargo tauri dev
# 4. Build the production installer
cargo tauri build
# Installer output: src-tauri\target\release\bundle\nsis\
Features
v0.2.0 — Available Now
| # | Feature | Description |
|---|---|---|
| 1 | Persistent Config | Settings saved to AppData\Roaming\VoiceTypeless\config.json and restored on relaunch |
| 2 | Persistent History | All recognition results saved to AppData\Roaming\VoiceTypeless\history.json |
| 3 | Clear All History | One-click clear with a confirmation dialog |
| 4 | Export History | Copy all transcription results to clipboard as formatted plain text |
| 5 | Search / Filter History | Real-time text search across all history items (frontend, no round-trip) |
| 6 | Statistics Panel | Total recordings, total characters, and per-language breakdown |
| 7 | Demo Mode | Simulate a full recording + transcription cycle — no microphone or speech model required |
| 8 | Theme Toggle | Dark, Light, and System theme options — persisted in config.json |
| 9 | Windows Autostart | Optional launch-at-startup via HKCU\...\Run registry key |
| 10 | System Tray | Minimize to tray with Show/Hide and Quit items in the tray context menu |
Future — Speech Integration
| Feature | Status |
|---|---|
| Global hotkey push-to-talk & free-speech | 🔜 In design |
| Offline speech recognition via SenseVoice | 🔜 Requires Go Core integration |
| Whisper-tiny model support | 🔜 Planned |
| GPU acceleration (DirectML / CUDA) | 🔜 Planned |
| Auto-paste transcription to active app | 🔜 Planned |
| Voice Activity Detection (VAD) auto-stop | 🔜 Planned |
| Plugin system (JS / Lua transform scripts) | 🔜 Planned |
| Windows 7 compatibility build | 🔜 Planned |
Architecture Overview
| Layer | Technology | Role |
|---|---|---|
| Desktop Shell | Tauri v2 (Rust) | Window management, IPC routing, OS integration, tray |
| Frontend | Svelte 5 + TailwindCSS + TypeScript | All user-visible UI; no business logic |
| Core Library | Go 1.23+ | Audio capture, speech recognition, text post-processing |
| Speech Engine | sherpa-onnx + ONNX Runtime | Offline inference (SenseVoice / Whisper) |
| Audio I/O | malgo (miniaudio Go bindings) | Microphone capture at 16 kHz mono |
The Go Core runs as a sidecar process spawned by Tauri and communicates via JSON-RPC 2.0
over a named pipe. For full design details see docs/architecture.md.
Configuration & Data Storage
VTL stores all user data in the Windows app-data folder:
%APPDATA%\Roaming\VoiceTypeless\
├── config.json # Application settings (hotkeys, theme, device, language, …)
└── history.json # Transcription history (capped at maxHistoryItems, default 50)
Default config is written automatically on first launch. To reset all settings, delete
config.json — it will be recreated with defaults on the next start.
API Reference
See docs/api.md for the complete Tauri command reference, including all
parameters, return types, and error codes.
Contributing
- Fork the repository and clone it locally
- Create a feature branch:
git checkout -b feat/your-feature - Follow the architecture design — keep the layer boundaries clean
- Test your changes with
cargo tauri dev - Open a pull request with a clear description of what changed and why
Please open an issue before beginning large or breaking changes so we can coordinate.
Changelog
See CHANGELOG.md for the full version history.
License
MIT © Voice-typeless Contributors
