Quickstart
From a fresh Mac to your first dictated sentence. The only slow part is the one-time model download, and the app tells you how long that will take.
1. Install the app
Download the DMG — one universal build for Apple Silicon and Intel, Developer-ID signed and Apple-notarized. Open it and drag OpenVoiceFlow into Applications.
Open the app from Applications, not from the disk image — macOS cannot launch an app off a mounted DMG. The first launch shows a “downloaded from the Internet” confirmation. That is normal for every notarized app outside the App Store; click Open. Nothing is broken.
2. Grant three permissions
Onboarding walks you through them one at a time, and each row unlocks only when the previous one is granted:
- Microphone — so the app can hear you. It listens only while the key is held.
- Accessibility — so it can paste at your cursor. It presses ⌘V on your behalf; it does not read your screen.
- Input Monitoring — so the push-to-talk key works in every app. One key is watched; every other keystroke passes straight through.
If you click Allow and nothing seems to happen, the screen has an Open System Settings escape hatch — full detail in Permissions.
3. Pick a speech engine
Four Whisper models are offered. One is marked RECOMMENDED, chosen from your Mac's chip and free disk space: Large turbo on Apple Silicon with room to spare, Small otherwise.
| Engine | Size | Best for |
|---|---|---|
| Tiny | 39 MB | Fastest. Fine for quick notes. |
| Small | 466 MB | Everyday dictation on any Mac. |
| Medium | 1.5 GB | Hears more, asks more of your Mac. |
| Large turbo | 1.6 GB | Hears the most. Best on Apple Silicon. |
The download runs once and shows size, speed, and an estimate. When the bar reaches the top it keeps working for a moment — that is macOS compiling the model for your Mac, and it only happens on first run. Details in Whisper models.
4. Say something
The last onboarding screen asks you to hold the key and talk. Words appear as you speak them. Release, and the text is inserted.
After that, the loop is the same everywhere: click into any text field, hold fn, speak, release.
By default macOS may open the emoji picker or Dictation when you press fn. Set System Settings ▸ Keyboard ▸ Press 🌐 to → Do Nothing, or pick a different key in Hotkeys.
What to do next
- Teach it the names it gets wrong — Personal dictionary (needs cleanup on).
- Make a spoken shortcut for text you type constantly — Snippets (works with cleanup off).
- Decide whether you want AI cleanup at all. It ships off, and off is a perfectly good place to stay.
Documentation for OpenVoiceFlow 0.5.5 · last updated 2026-07-31. Something wrong or missing here? That is a bug — tell us at shimoverse@gmail.com.