Install

Requirements

• A Mac with Apple Silicon (M1 or later). Intel Macs aren’t supported, because the models run on Apple’s MLX framework.
• About 2.5 GB of free disk space for the default models. Kass checks for the model sizes plus 500 MB of headroom before it downloads.
• An internet connection for the first model download. After that, Kass works offline.

Download and install

1. Go to the download page. The newest .dmg starts downloading from GitHub on its own. If it doesn’t, use the link on that page or grab it from the Releases page.
2. Open the .dmg from your Downloads folder and drag Kass into Applications.
3. Open Kass from Applications.

Keep it in Applications

macOS only keeps permissions like Accessibility and Input Monitoring for apps in /Applications. If you open Kass from somewhere else, it offers to Move and relaunch for you.

Release builds are signed and notarized by Apple, so they open like any other Mac app.

First-run setup

Onboarding walks you through everything. It usually takes a few minutes, most of it the model download, which keeps going in the background while you continue.

1. Download models. Kass downloads Whisper Turbo (about 1.6 GB) for speech and Qwen3 0.6B (about 400 MB) for cleanup. A failed download retries on its own. If it keeps failing, you can try again or pick a smaller speech model.
2. Input Monitoring. This lets Kass notice your hotkey chord from any app. macOS asks you to quit Kass after you allow it. When Kass reopens, onboarding comes back to the same step and the download picks up where it left off.
3. Accessibility. This lets Kass put text into the app you’re using. No restart needed. Without it, your dictations are still saved in Captures, but they won’t be typed for you.
4. Microphone. Allow access, say hello, and pick a different microphone if you like.
5. Keys. Learn the dictation chord. You can also choose to open Kass when you log in.
6. Your name. Say your name and fix its spelling. It’s saved to your dictionary.
7. Practice. Try some messy speech and a Command Mode rewrite. These steps unlock once the models have finished downloading.

Any permission step can be skipped and granted later. You can run onboarding again from Settings › Features or from the command palette (⌘ K › Open onboarding).

Updating

Kass tells you when a new version is out. Get the newest .dmg from the download page and replace the copy in Applications. Your captures, dictionary, styles and models are kept.

Updates are signed with the same Developer ID, so Kass keeps its Microphone, Accessibility and Input Monitoring permissions.

Coming from Herga or Voicebox

Kass used to be called Herga, and before that Voicebox. The first time Kass opens, it brings over your captures, writing styles, dictionary and settings. macOS sees Kass as a new app, so it asks for Microphone, Accessibility and Input Monitoring once more.

If Herga updated itself to Kass, there’s nothing else to do. If you installed Kass alongside Herga, Kass moves Herga.app to the Trash so it doesn’t start at login and answer the same hotkey. An old Voicebox.app you move to the Trash yourself.

Uninstalling

Quit Kass and move it from Applications to the Trash. Your captures and settings stay in Kass’s data folder, which you can open from Settings › General. Your models stay in the Hugging Face cache (or wherever you moved them in the Models tab). Delete those too if you want the space back.