Offline dictation for macOS

Push-to-talk: hold left Option and speak. Release, and LocalFlow transcribes on your Mac with Whisper and pastes the text into the app in front. No audio leaves the machine.

Open source under the MIT License. Linux and Windows: best-effort.

How LocalFlow works, in 3 steps

  1. Hold the Option key

    Hold left Option, the default hotkey, in any app. A combo hotkey toggles recording instead.

  2. Dictate while LocalFlow listens

    The recording pill shows LocalFlow is listening. Audio and recognition stay on your machine.

  3. Release to paste the dictation

    Whisper transcribes, optional cleanup fixes punctuation, and the text pastes at your cursor in the app in front.

LocalFlow keeps your voice on your machine

Speech-to-text runs offline with Whisper through faster-whisper. Audio and recognition stay on the computer, and LocalFlow makes no cloud calls.

  • Optional cleanup runs on a local Ollama model
  • Phone access goes through your own Tailscale network

Platforms LocalFlow supports

A native setup wizard picks and downloads the speech and cleanup models, and sets the microphone and hotkey.

LocalFlow support by platform
PlatformDictation and pasteMeetings
macOSFull supportSupported
LinuxBest-effortDisabled
WindowsBest-effortDisabled

Download LocalFlow

Paste one command into a terminal. The installer sets up its own Python environment and opens the setup window.

  • On macOS, Homebrew, git and Python are optional.
  • The macOS app is unsigned: on first launch, right-click it and choose Open, or use Open Anyway in System Settings, Privacy & Security.
macOS and Linux
curl -fsSL https://raw.githubusercontent.com/CortexRND/localflow/main/install.sh | bash

Windows (PowerShell)
irm https://raw.githubusercontent.com/CortexRND/localflow/main/install.ps1 | iex

macOS and Linux, with Ollama cleanup
curl -fsSL https://raw.githubusercontent.com/CortexRND/localflow/main/install.sh | bash -s -- --with-ollama