whatisgithub

What is macintalk?

leonardm01/macintalk — explained in plain English

Analysis updated 2026-05-18

1SwiftAudience · generalComplexity · 3/5Setup · moderate

In one sentence

A macOS menu bar app that transcribes your speech on device with a hotkey, cleans it up using Apple Intelligence, and pastes the polished text wherever you were typing.

Mindmap

mindmap
  root((repo))
    What it does
      On device speech transcription
      AI text cleanup
      System wide paste
    Tech stack
      Swift
      SpeechAnalyzer
      Foundation Models
      Xcode
    Use cases
      Dictate text in any app
      Clean up rambling speech
      Review past transcriptions
    Requirements
      Apple silicon Mac
      Apple Intelligence enabled
      Microphone access
    Audience
      Mac power users
      Developers

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

What do people build with it?

USE CASE 1

Dictate text with a hotkey and have it pasted, cleaned up, into any Mac app you are using.

USE CASE 2

Review a local history of your past voice transcriptions in both raw and cleaned form.

USE CASE 3

Choose a cleanup tone, casual, balanced, or business, to match how the transcribed text sounds.

What is it built with?

SwiftSpeechAnalyzerFoundation ModelsXcode

How does it compare?

leonardm01/macintalkaakarsh-goyal/macjuicealtuzar/sonicflow
Stars111
LanguageSwiftSwiftSwift
Setup difficultymoderateeasyeasy
Complexity3/51/52/5
Audiencegeneralgeneralgeneral

Figures from each repo's GitHub metadata at analysis time.

How do you get it running?

Difficulty · moderate Time to first run · 30min

Requires an Apple silicon Mac with a recent macOS and Xcode version, plus Apple Intelligence enabled.

So what is it?

MacinTalk is a small menu bar utility for Mac that lets you dictate text anywhere on your system using your voice. It lives quietly in your menu bar and transcribes your speech directly on your device using Apple's SpeechAnalyzer technology, then cleans up the resulting text with Apple Intelligence before pasting the polished result into whatever app you were typing in. To use it, you hold a hotkey, Control plus Option plus Space, speak your sentence, and let go. MacinTalk transcribes what you said locally on your Mac, tidies up filler words and grammar using Apple's on device Foundation Models when they are available, and then pastes the cleaned text into the app you had focused, using the clipboard and a simulated paste keystroke. If Apple Intelligence is not available on your machine, it simply pastes the raw, uncleaned transcript instead. The main window has a Home section where you can watch your dictation happen live, see a partial transcript as you speak, and browse a locally stored history of previous transcriptions in both their raw and cleaned forms, with options to copy, delete, or clear that history. A Settings section lets you pick a cleanup style, casual, balanced, or business, and manage the app's permissions and any speech recognition assets it needs installed. Because this is an early, MVP stage tool, it only works on macOS, not iOS, and it depends on Apple's speech assets being installed for your language before it can work fully offline. It temporarily uses your clipboard during the paste step, restoring your previous clipboard contents when possible, and it does not yet support cloud fallback or custom vocabulary. Running it requires an Apple silicon Mac with a fairly recent version of macOS and Xcode, Apple Intelligence turned on in system settings, and granting the app microphone access along with input monitoring and accessibility permissions so it can listen for the hotkey and paste text system wide. Internally, the app is organized around a coordinator that manages the record, clean, and insert steps, a service for live microphone transcription, a cleaner for on device text polishing, a global hotkey listener, and a component that safely handles clipboard based text insertion.

Copy-paste prompts

Prompt 1
Help me build and run MacinTalk in Xcode 26 and grant it the permissions it needs to work.
Prompt 2
Explain how MacinTalk cleans up dictated speech using Apple's on device Foundation Models.
Prompt 3
Add a cloud fallback option to MacinTalk for when on device speech assets are not installed.
Prompt 4
Add support for custom vocabulary to MacinTalk so it transcribes specific names and terms correctly.

Frequently asked questions

What is macintalk?

A macOS menu bar app that transcribes your speech on device with a hotkey, cleans it up using Apple Intelligence, and pastes the polished text wherever you were typing.

What language is macintalk written in?

Mainly Swift. The stack also includes Swift, SpeechAnalyzer, Foundation Models.

How hard is macintalk to set up?

Setup difficulty is rated moderate, with roughly 30min to a first successful run.

Who is macintalk for?

Mainly general.

Open on GitHub → Ask about another repo

This repo across BitVibe Labs

Verify against the repo before relying on details.