matthartman/ghost-pepper: Hold-to-talk speech-to-text for macOS. 100% local, powered by WhisperKit and local LLM cleanup. Hold Control to record, release to transcribe and paste. · GitHub

100% local hold-to-talk speech-to-text for macOS. Hold Control to record, release to transcribe and paste. No cloud APIs, no data leaves your machine.

Download the latest release — macOS 14.0+, Apple Silicon (M1+)

• Hold Control to talk — release to transcribe and paste into any text field
• Runs entirely on your Mac — models run locally via Apple Silicon, nothing is sent anywhere
• Smart cleanup — local LLM removes filler words and handles self-corrections
• Menu bar app — lives in your menu bar, no dock icon, launches at login
• Customizable — edit the cleanup prompt, pick your mic, toggle features on/off

Ghost Pepper uses two open-source models that download automatically on first launch:

Models are served by Hugging Face and cached locally after first download.

Download the app:

1. Download GhostPepper.dmg
2. Open the DMG, drag Ghost Pepper to Applications
3. Grant Microphone and Accessibility permissions when prompted
4. Hold Control and speak

Build from source:

1. Clone the repo
2. Open GhostPepper.xcodeproj in Xcode
3. Build and run (Cmd+R)

• Launch at login is enabled by default on first run. You can toggle it off in Settings.
• No logging to disk — transcriptions are never written to files. Debug logs are in-memory only and disappear when the app quits.

Built with WhisperKit, LLM.swift, Hugging Face, and Sparkle.

MIT

All models run locally, no private data leaves your computer. And it’s spicy to offer something for free that other apps have raised $80M to build.

Ghost Pepper requires Accessibility permission, which normally needs admin access to grant. On managed devices, IT admins can pre-approve this via an MDM profile (Jamf, Kandji, Mosaic, etc.) using a Privacy Preferences Policy Control (PPPC) payload: