Live Korean speech transcription/translation overlay app. Captures microphone audio, detects phrases via voice activity detection, sends them to an OpenAI-compatible Whisper API, optionally translates via a Chat API, and displays results in a frameless always-on-top window.
- Real-time speech transcription using Whisper API
- Optional translation via Chat completions API (with scientific terminology preservation)
- Configurable via
settings.yml(created next to the binary on first run) - Separate settings window with API URLs, API keys, language selection, and more
- Frameless, draggable, resizable overlay window
- Supports 10 source languages (transcription) and 50+ target languages (translation)
- Rust (1.85+ for edition 2024)
- An OpenAI-compatible Whisper API endpoint
sudo apt-get install -y pkg-config libasound2-dev libssl-dev \
libgtk-3-dev libxcb-render0-dev libxcb-shape0-dev libxcb-xfixes0-dev libxkbcommon-devNo extra dependencies. CoreAudio and Security Framework ship with the OS.
No extra dependencies. WASAPI and SChannel are built-in.
cargo build --releaseBinary outputs to target/release/livecaptran (or livecaptran.exe on Windows).
On first run, a settings.yml file is created next to the binary with default settings (OpenAI API endpoints). Edit this file or use the in-app settings window (gear icon) to configure:
- Transcribe API URL / Key - Whisper-compatible transcription endpoint
- Chat API URL / Key / Model - Chat completions endpoint for translation
- Source language - Language being spoken
- Target language - Translation target (or "None" to disable)
- Display mode - Show both transcription + translation, or translation only
- Font size and VAD sensitivity
GitHub Actions builds for all three platforms on push to main. See .github/workflows/build.yml. Download artifacts from the Actions tab.
| Platform | Target | Binary |
|---|---|---|
| Linux | x86_64-unknown-linux-gnu |
livecaptran |
| Windows | x86_64-pc-windows-msvc |
livecaptran.exe |
| macOS | aarch64-apple-darwin |
livecaptran |