OpenVoiceFlow documentation

Everything about running OpenVoiceFlow on your Mac — from first launch to the exact bytes that do and don't leave your machine. Written against the shipping app, version 0.5.27.

OpenVoiceFlow is a free and open source push-to-talk dictation app for macOS. You hold a key, speak, release, and cleaned text lands at your cursor in whatever app you were using. Whisper runs on your Mac, so your audio never leaves it.

New here?

Two minutes gets you dictating: Quickstart. If you haven't downloaded the app yet, start at the download page.

Start here

QuickstartDownload to first dictated sentence, in about two minutes. InstallationThe DMG, the Gatekeeper prompt, and the macOS 12–13 fallback build. PermissionsThe three macOS permissions, what each one buys, and how to fix a stuck grant. Whisper modelsWhich of the four engines to pick for your Mac, and what changes between them.

What OpenVoiceFlow does

Capability	Where it runs	Details
Push-to-talk dictation	Your Mac	Hold a key, speak, release — works in any app with a text cursor.
Speech-to-text	Your Mac	On-device Whisper via WhisperKit, four model sizes.
Personal dictionary	Your Mac + cleanup	Names and jargon it should never misspell.
Snippets	Your Mac	Say a short trigger, get the long text. No network call.
Per-app styles	Cleanup	Casual in Slack, formal in Mail.
AI cleanup	Off by default	Optional — your own API key, or a fully local model.

What it does not do

An honest manual saves you the trouble of looking for something that isn't there. As of 0.5.27:

macOS only. macOS 14 (Sonoma) or newer, Apple Silicon and Intel. There is no Windows or Linux build. iOS and Android are in development and cannot be downloaded today.
No spoken punctuation commands. Saying “comma” types the word “comma”. Whisper punctuates from sentence structure on its own, and AI cleanup handles the rest if you turn it on.
No meeting recording, no cloud sync, no accounts. There is no server your dictation history syncs to.
Dictation itself is never reported anywhere. As of 0.5.7, an opt-out anonymous usage summary (word/time totals, which features you use, a display name you choose) powers an in-app leaderboard — never audio, never dictated text. Off in one toggle: details and how to turn it off.

The whole manual

Dictation basicsThe hold-and-speak loop, the HUD, limits, and what each state means. Improving accuracyThe changes that actually move the needle, in order of effect. Settings referenceEvery setting, its default, and what it changes. Privacy architectureEvery file written and every byte that can leave your Mac. TroubleshootingSymptoms, causes, and fixes for the problems people actually hit. FAQShort answers to the questions we get most.

Documentation for OpenVoiceFlow 0.5.27 · last updated 2026-09-17. Something wrong or missing here? That is a bug — tell us at contact@openvoiceflow.com.