Install YapToText
The simplest way to install YapToText is the Mac App Store. It is free, the speech model and the cleanup model are inside the app, and it works offline from the first launch. If you would rather use Homebrew, run brew install --cask ryleighnewman/yaptotext/yaptotext. That build is about 18 MB, and you download the speech model after the first launch.
Either way you need macOS 14 (Sonoma) or later on Apple Silicon, and the app will ask for two permissions when it first runs: Microphone, which it cannot work without, and Accessibility, which is what lets it type into other apps.
Mac App Store
The App Store build is the one I recommend. It costs nothing, and there is no paid tier, no locked feature, and no account to make. The only thing in the app that costs money is a tip jar you can ignore.
Both models ship inside the app: Whisper Large v3 Turbo (Q5) for speech and Phi-3.5 Mini for cleanup. Nothing has to be downloaded, so the first dictation works with the network off. That is also why the app is big, about 2.8 GB with both models inside.
Updates come through the App Store. The app never checks a server on its own.
Homebrew
Homebrew support arrived in 1.3.1. The cask lives in my own tap, so one command installs it:
- brew install --cask ryleighnewman/yaptotext/yaptotext
The Homebrew build is the same app, built from the same source, at the same version as the App Store one. There are two differences. It is signed with a Developer ID certificate and notarized instead of App Store signed, so Gatekeeper opens it normally. And it ships without the speech and cleanup models. With them the app is 2.8 GB, which is past the 2 GB limit for a GitHub release file; without them it is about 18 MB.
So after the first launch, open the AI Models page and download a speech model. Whisper Large v3 Turbo (Q5) is the recommended one at 574 MB. Raw Transcription needs only the speech model. Every other built-in mode runs the cleanup model too, so if you want those, download Phi-3.5 Mini Instruct as well; it is about 2.3 GB. Downloads come over HTTPS from Hugging Face, are stored in the app's own container, survive updates, and can be removed any time. Those downloads are the only network traffic the app can generate, and only when you start them.
Two things to know. The cask installs YapToText.app into /Applications, so it replaces a Mac App Store copy if one is there; pick one install method. And the first Homebrew build, 1.3.1, could not use the microphone. 1.4 fixed that, so if you installed earlier, run brew upgrade --cask yaptotext.
Requirements
- macOS 14 (Sonoma) or later. On macOS 26 the interface picks up the Liquid Glass look.
- Apple Silicon.
- A microphone, since the app has to hear you.
The Homebrew cask checks the macOS version for you and refuses to install on anything older than Sonoma.
First-run permissions
The first launch walks you through the two permissions.
- Microphone is required, so the app can hear you. macOS asks once, and after that dictation starts instantly. If a dictation ever reports that microphone access is needed, turn YapToText on under System Settings, Privacy & Security, Microphone.
- Accessibility is what lets YapToText type into other apps and read the text around your cursor. macOS only lets apps with that permission type into other apps. Without it, YapToText still transcribes everything and copies the result to your clipboard, but reading selected text for Regenerate is unavailable.
An app cannot grant itself Accessibility. YapToText opens the right pane in System Settings, Privacy & Security, Accessibility, and you flip the switch next to its name yourself.
Your first dictation
- Click where you want the text to go: an email, a note, a chat box.
- Tap Right Command once. The floating panel appears and the menu bar capybara's speech bubble turns red.
- Say what you want to write. The live text forms in the panel as you talk.
- Tap Right Command again. Your words are typed at the cursor.
While dictating, Space pauses and resumes, Esc cancels without typing or saving anything, and 1 through 9 switch the output mode. The first dictation after a launch loads the model, which takes a moment; every dictation after that is quick.
Building from source
The whole app is on GitHub under GPL-3.0. Clone it, open YapToText.xcodeproj, select the YapToText scheme, and press Command-R. You need Xcode 26 or later. The app is sandboxed and builds the same way it ships.
If something in the install does not go the way this page says, tell me on the issues page.
Questions
Is the Homebrew version of YapToText the same as the App Store version?
Yes. It is built from the same source at the same version. The only differences are that it is Developer ID signed and notarized instead of App Store signed, and the speech and cleanup models are not inside it, so you download them from the AI Models page.
Which model should I download after installing YapToText with Homebrew?
Whisper Large v3 Turbo (Q5), the recommended speech model, at 574 MB. Raw Transcription needs only that. Every other built-in mode runs the cleanup model too, so add Phi-3.5 Mini Instruct, about 2.3 GB, if you want them.
Does YapToText need an internet connection?
The App Store build does not. Both models are inside the app, so it works offline from the first launch. The Homebrew build needs a connection once, for the model download you start yourself. There are no other network calls.
Can I have the App Store and Homebrew versions installed at the same time?
No. The Homebrew cask installs YapToText.app into /Applications and replaces an App Store copy that is already there. Choose one install method.
How do I update YapToText?
The App Store build updates through the Mac App Store. The Homebrew build updates with brew upgrade --cask yaptotext. The app never checks for updates on its own.