Skip to content
Kass

Kass docs

Kass is a dictation app for Apple Silicon Macs. Hold a key chord in any app and speak. When you let go, Kass transcribes what you said, cleans it up the way you’d have typed it, and puts the text into the field you were typing in.

Everything happens on your Mac. Speech recognition runs on Whisper, cleanup runs on a small local language model (Qwen3), and both use Apple’s MLX framework on the GPU. There’s no account, no server and no subscription.

Kass is named for the bard in Breath of the Wild who carries songs from place to place, the way Kass carries your words into whatever app you’re in. It used to be called Herga, and before that Voicebox. If you’re coming from either, see Coming from Herga or Voicebox.

What cleanup does, and doesn’t do

Section titled “What cleanup does, and doesn’t do”

Speech comes out with filler words, false starts and changes of mind that typed text wouldn’t have. Cleanup:

  • removes filler words (“um”, “uh”, “like”), stutters and repeats
  • applies spoken self-corrections: “Tuesday, no actually Wednesday” becomes “Wednesday”
  • adds punctuation and capitals in your writing style
  • follows spoken commands for line breaks, lists and symbols
  • spells names and terms the way your dictionary says

It keeps your words and your meaning. It never summarizes and never adds anything you didn’t say. When you do want text rewritten, use Command Mode.

Where to start

Section titled “Where to start”