Press your hotkey, talk, and clean text lands wherever your cursor is. Transcription runs locally with Whisper; optional LLM cleanup fixes grammar and trims filler. Runs on macOS, Linux, and Windows.

• Local transcription — Whisper (embedded whisper.cpp + a GGML model) runs on-device. Your audio never leaves your machine.
• Optional cleanup — fix grammar and drop fillers via Ollama, LM Studio, OpenAI, Anthropic, OpenRouter, or managed MegaBrain Cloud. Five correction levels and a tone setting.
• Works everywhere you type — pastes into any app via the clipboard.
• Voice snippets — say a trigger, get a full expansion.
• Voice commands — map spoken phrases to keyboard shortcuts.
• Configurable hotkey — any accelerator, or a bare modifier (e.g. Control). Push-to-talk or toggle.
• Privacy mode — keep everything local and stop saving history.
• Auto-updates — signed over-the-air updates.

Download the installer for your platform from the latest release, or grab it from dict.studio. Existing installs update themselves over the air.

Whisper is built in (whisper.cpp is embedded, Metal-accelerated on Apple Silicon), so there's nothing else to install. You just need a model: the in-app model manager under Preferences → Speech downloads a GGML model for you.

Requires Rust and the Tauri CLI.

# Rust
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
cargo install tauri-cli --version "^2"

cd src-tauri
cargo check          # quick type-check
cargo test           # unit tests
cargo tauri dev      # hot-reload dev build
cargo tauri build    # release bundle

Hotkey → record (cpal) → Whisper (embedded whisper.cpp) → optional LLM cleanup → paste (clipboard + key sim)

Built with Rust + Tauri 2.0; the frontend is vanilla HTML/CSS/JS (no build step). See CLAUDE.md for the full architecture and CONFIG.md for every config key.

• macOS 14+ / Linux (GTK3, WebKitGTK) / Windows 10+
• Microphone permission (prompted automatically)
• Accessibility permission on macOS (for pasting into other apps)
• A GGML model (downloaded in-app; whisper.cpp is embedded, no separate install)
• An OpenAI-compatible endpoint or a MegaBrain Cloud account if you enable LLM cleanup

English and Brazilian Portuguese today (whisperLanguage maps to Whisper's -l flag; auto lets Whisper detect).

MIT