Offline voice-to-text input tool. Hold Right Ctrl, speak, release — text copied to clipboard.

中文文档

Hold Right Ctrl → Speak → Release → Paste anywhere

• Local Inference — Whisper AI model runs entirely on your machine. After first-launch model download, no internet needed and no data ever leaves your device.
• Lightweight Binary — Small executable (~15MB), download and run. The ~466MB model is fetched automatically on first launch (with resume support).
• Global Hotkey — Works in any application. No window switching needed.
• Cross-Platform — Linux, Windows, macOS.
• Real-time Waveform — Visual feedback while recording.
• Chinese Optimized — Tuned for Chinese speech with punctuation post-processing. English works too.
• System Tray — Runs silently in the background. No taskbar clutter.
• Resume Downloads — Model download supports resume from interruption.

Hold Right Ctrl  →  Speak  →  Release Right Ctrl  →  Ctrl+V to paste
     ┌────────────────────────────────────┐
     │  Recording... (waveform shown)     │
     │  "正在聆听..."                      │
     └────────────────────────────────────┘
              ↓ release key
     ┌────────────────────────────────────┐
     │  Transcribing...                   │
     └────────────────────────────────────┘
              ↓ done
     ┌────────────────────────────────────┐
     │  你好世界这是语音输入测试            │
     │  Copied to clipboard               │
     └────────────────────────────────────┘

Go to Releases and download the latest binary for your platform:

# Linux / macOS — make executable and run
chmod +x voiceinput
./voiceinput

# Windows — just double-click
voiceinput.exe

On first launch, the Whisper model (~466MB) is automatically downloaded. The download supports resume — if interrupted, it will continue from where it left off on next launch.

• China users: Model is downloaded from hf-mirror.com (auto-detected from system locale).
• Other regions: Model is downloaded from huggingface.co.
• Manual override: Set mirror = "cn" or mirror = "global" in config file (see below).

1. A microphone icon appears in the system tray
2. Hold Right Ctrl — overlay appears, start speaking
3. Release Right Ctrl — transcription begins
4. Result is copied to clipboard — Ctrl+V to paste anywhere

Config file is auto-created at:

• Linux: ~/.config/voiceinput/config.toml
• Windows: %APPDATA%\voiceinput\config.toml
• macOS: ~/Library/Application Support/voiceinput/config.toml

[hotkey]
trigger = "RControl"       # Trigger key
min_hold_ms = 300          # Minimum hold to activate (ms)

[audio]
sample_rate = 16000        # Hz (Whisper requirement)
max_duration_sec = 60      # Auto-stop after this
min_duration_ms = 500      # Ignore if shorter

[whisper]
language = "zh"            # "zh", "en", "ja", etc.
beam_size = 5              # Beam search width
threads = 0                # 0 = auto (CPU cores - 1)

[general]
log_level = "info"         # debug, info, warn, error
mirror = "auto"            # Model download: "auto", "cn", "global"

Linux extras:

sudo apt install libwebkit2gtk-4.1-dev libappindicator3-dev \
  librsvg2-dev libasound2-dev libxdo-dev

Windows extras: Visual Studio Build Tools with "Desktop C++" workload.

git clone https://github.com/YOUR_USERNAME/voiceinput.git
cd voiceinput

# Dev build (model auto-downloads on first run)
cd src-tauri && cargo build

# Release build
cd src-tauri && cargo build --release

Push a tag to auto-build for all platforms:

git tag v1.0.0
git push origin v1.0.0
# → GitHub Actions builds Linux/Windows/macOS
# → Artifacts published to GitHub Releases

You can also trigger builds manually from the Actions page.

MIT