Local, cross-platform, open-source speech-to-text dictation app. A privacy-first alternative to SuperWhisper and WisprFlow.
Calliope runs Whisper inference entirely on your machine — no cloud, no API keys required. Just press a hotkey and dictate.
- Fully local — all transcription happens on-device via whisper.cpp
- Cross-platform — macOS, Windows, and Linux
- Push-to-talk & toggle mode — with automatic silence detection
- Text injection — transcribed text is typed directly into the focused app
- Optional post-processing — clean up transcriptions with a local or cloud LLM (Ollama, LM Studio, OpenRouter)
- GPU accelerated — Metal (macOS), CUDA (NVIDIA), ROCm/HIPBlas (AMD)
- Lightweight — lives in your system tray
- Tauri v2 (Rust + WebView)
- whisper-rs (Rust bindings to whisper.cpp)
- React 19 + TypeScript + Tailwind CSS v4
- Rust
- Bun
- Platform-specific dependencies — see Tauri prerequisites
bun install
bun run tauri devTo build with Whisper inference enabled:
bun run tauri dev -- --features whisper # CPU only
bun run tauri dev -- --features metal # macOS (Apple Silicon / Intel)
bun run tauri dev -- --features cuda # NVIDIA GPU
bun run tauri dev -- --features rocm # AMD GPU (ROCm/HIPBlas)bun run tauri build --features metal # macOS releaseMIT