Analysis updated 2026-05-18
Dictate text into any Mac app and have AI clean it up into polished writing.
Turn spoken notes into a formatted email, report, or list of action items.
Select text in any app and speak a command like summarize or translate.
Build a personal dictionary of technical terms and names for more accurate recognition.
| punk2898/punktype | ronaldo-avalos/maya | aymandakir-gh/gh-notch | |
|---|---|---|---|
| Stars | 27 | 27 | 28 |
| Language | Swift | Swift | Swift |
| Setup difficulty | easy | moderate | easy |
| Complexity | 2/5 | 2/5 | 2/5 |
| Audience | general | general | vibe coder |
Figures from each repo's GitHub metadata at analysis time.
Requires macOS 14 or later and your own API key from DeepSeek or another compatible provider.
PunkType is a macOS menu bar app that lets you speak instead of type. You press a keyboard shortcut, say what you want to write, press the shortcut again to stop, and the text is automatically cleaned up by AI and pasted wherever your cursor is. It runs in the background as a small icon in the menu bar and requires macOS 14 or later. The app works in two modes. In standard dictation mode, your voice is converted to text by your Mac's built-in speech recognition, then sent to DeepSeek (an AI service you connect with your own API key) to turn spoken language into polished written text. There are three output styles: a fast mode that types the raw transcription with no AI processing, a polished mode that removes filler words and cleans up informal speech, and a formatted mode that restructures your words into something like an email, a report, a summary, or a list of action items. In command mode, you first select some text in any app, then press the shortcut and speak an instruction, such as "summarize this" or "translate to English." The result appears in a popup where you can copy it or replace the selected text directly. This works across any application on your Mac. The app also builds a personal dictionary from your usage, extracting technical terms, names, and product names to help the recognition avoid errors on words specific to your work. You can also enable OpenAI's cloud transcription service as a fallback if the local recognition is inaccurate. All data goes directly from your machine to the AI provider using your own API key, with no intermediate server involved. The README is written in Chinese with a linked English version. The project is released under the MIT license and the code is auditable and compilable from source.
A macOS menu bar app that turns your speech into cleaned-up typed text using AI, or runs voice commands on selected text.
Mainly Swift. The stack also includes Swift, SwiftUI, AVFoundation.
Released under the MIT license, so anyone can use, modify, and share the code freely, including for commercial purposes.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly general.
This repo across BitVibe Labs
Verify against the repo before relying on details.