MacOS transcribe audio using FFmpeg and OpenAI Whisper with Automator Quick Action and Keyboard Shortcut

This project provides a script to transcribe audio using FFmpeg and OpenAI Whisper. It includes instructions for setting up a macOS Automator Quick Action to record audio from the default microphone, transcribe it, and paste the transcribed text into the current focused input. The script handles recording, stopping the recording, transcribing the audio, and managing the output files

Install

• brew install ffmpg
• brew install openai-whisper

or compile from the source

Create automator Quick Action

• add Run Shell Script action
• add: osascript <full path to the project>/transcribe.scpt
• Save as Transcribe

Go to System Preferences

• Security & Privacy -> Privacy -> Accessibility - Add Automator
• Keyboard -> Shortcuts -> Services -> Add Shortcut for Transcribe

Debug

• First of all, check if the ffmpeg is around version 7 :-)
• Check if it records the correct audio stream. I use avfoundation -i :1 ([1] MacBook Pro Microphone).
  • Try ffmpeg -f avfoundation -list_devices true -i ""
• Try if the whisper works well. If you don't have the model downloaded, whisper will download it for you. eg. turbo is around 6GB!