Quickstart
Install the macOS service, build the UI, and start dictating.
Installation
Choose the Homebrew installer, manual Homebrew setup, or source setup path.
Mobile apps
Understand the Flutter app, iOS keyboard, Android input method, and model packs.
Configuration
Review the full TOML reference for engines, grammar, audio, output, and TTS.
Architecture
See how Python, Swift, Flutter, native mobile, IPC, and local models fit together.
Runtime paths
| Runtime path | Where it runs |
|---|---|
| Recording and audio cleanup | On device |
| Parakeet-TDT v3 and Qwen3-ASR transcription | In-process MLX |
| Android Parakeet-TDT v3 and Qwen3-ASR transcription | On-device sherpa-onnx |
| WhisperKit transcription | Localhost |
| Grammar cleanup | Apple Intelligence on-device; Ollama or LM Studio on localhost or private LAN |
| Text-to-speech | In-process Kokoro MLX |
| History and backups | ~/.whisper/ |
Surfaces
| Surface | What it provides |
|---|---|
| macOS service and CLI | Global dictation hotkey, local transcription, grammar cleanup, replacements, clipboard or paste output, history, TTS, and service control through wh. |
| macOS Swift UI | Menu bar status, engine and grammar switching, overlay states, history, settings, updates, and service controls. |
| Flutter iOS app and keyboard | Local WhisperKit/Core ML transcription, setup, model packs, local history, modes, punctuation, haptics, and text-field workflows. |
| Flutter Android app and keyboard | Local WAV recording, sherpa-onnx transcription, model packs, keyboard verification, punctuation, quick insert, and haptics. |
Project promise
Local Whisper is built for local dictation and local text handling. Setup, model downloads,wh update, and repair commands can use the network. Normal audio and transcript processing stays on-device, localhost, or a private LAN server you explicitly configure.