Never leaves your Mac
Speech recognition runs on the Neural Engine. Your audio and text stay on your Mac, offline or not.
On-device dictation for Mac
Press fn, say what you want to write, and press it again. Your words land in whatever app you’re using. Transcription happens on your Mac, so your voice never leaves it.
Version 1.47.0, 10 MB.
Needs a Mac with Apple silicon and macOS 15 or later.
Standup notes
No account, no cloud, nothing to set up. Quack sits in your menu bar and waits for you to hold a key.
Speech recognition runs on the Neural Engine. Your audio and text stay on your Mac, offline or not.
Words appear in a pill at the bottom of your screen while you speak. When you press fn again, the finished text lands in the app you’re using.
Say Send it
to press Return, or Trash what I said
to throw the dictation away. Change your mind mid-sentence with sorry
or I mean
, and Quack types only the correction. Esc cancels.
Control API
Other programs can start Quack, stop it, and take the words.
Dictionary API
Point an agent at a folder. It adds the names Quack would get wrong.
import { createClient } from "@trpc/client" import { invoke } from "@tauri-apps/api" import { create } from "zustand" // Rolled out with kubectl once the // pgvector migration has landed. export const deploy = async () => { … }
quackctl add kubectl --heard-as "cube control"
The file is in your Downloads folder.
macOS asks whether you want to open an app downloaded from the internet. Choose Open. It only asks once.
Quack asks for both. The microphone lets it hear you, and Accessibility lets it type into other apps.
The first time it opens, Quack downloads its speech model, about 215 MB, and shows the progress at the top of your screen. When it says Quack is ready, you can start talking.
Send it
Trash what I said
Monday, sorry, Thursday
No wait,
I mean, and
actually make thatwork too, and
scratch thatdrops the sentence before it.