Polly is a blazing-fast, native macOS menu bar app that uses the Groq API (whisper-large-v3) to provide instant, highly accurate voice dictation right where your cursor is.
- Blazing Fast: Powered by Groq's high-speed Whisper endpoints.
- Native macOS: Built purely in Swift, lives as a lightweight agent in your menu bar.
- Hold-to-Talk: Simply press and hold the
fn(Globe) key to speak, and release to instantly paste the text into whatever app you're using. - Floating UI: A beautiful, non-intrusive floating badge lets you know when it's listening.
- Language Lock: Restrict dictation to a specific ISO language code (e.g.
en,ko) to prevent cross-language hallucinations. - Auto-Launch: Optionally start automatically when you log in.
No bulky Xcode project required! Polly compiles directly via the Command Line Tools.
- Clone the repository and navigate to the folder.
- Build the app:
This script compiles the Swift files, copies the icon templates, and outputs the final
chmod +x build.sh ./build.sh
Polly.appbundle right in the same directory! - Launch the app:
open Polly.app
When you first run Polly, you'll see its parrot mask icon appear in the menu bar.
- Accessibility Permission: Polly needs Accessibility rights to synthesize the
Cmd+Vkeystroke so it can paste your text anywhere. Go to System Settings > Privacy & Security > Accessibility and ensure Polly is toggled on. - Microphone Permission: macOS will ask for mic access the first time you attempt to record.
- Click the parrot icon in the menu bar and select Settings....
- Enter your Groq API Key (you can get one from the Groq console).
- Test your key to verify connectivity.
- Enter your 2-letter ISO language code (e.g.,
enfor English). - Toggle "Launch on system startup" if desired.
- Click any text box in any app.
- Press and hold the
fn(Globe) key. - A floating indicator will appear. Speak naturally!
- Release the
fnkey. - Within milliseconds, Polly will paste the transcibed text directly where your cursor was!