Whisperer

Whisperer is a Python application that records audio when a specific key is held down, and sends the audio to OpenAI's Whisper ASR system for transcription when another key is tapped. It then types the transcription into the active window.

Prerequisites

• Python 3
• OpenAI API key

Dependencies

• sounddevice
• numpy
• openai
• pynput
• scipy
• pyperclip

You can install these dependencies using pip:

pip install sounddevice numpy openai pynput scipy pyperclip

On Linux, also run:

sudo apt-get install xclip

Setup

1. Clone the repository.
2. Create a file named openai_api_key.txt in the root directory of the project.
3. Paste your OpenAI API key into openai_api_key.txt.

Usage

Run whisperer.py to start the application.

python whisperer.py

To quit the application, press ctrl + c.

• Hold down right ctrl button to start recording audio.
• Release right ctrl button to stop recording audio.

If you want to translate the recorded audio to French, tap the right shift button while recording

Notes

• The audio is recorded at a sample rate of 16000 Hz and saved as output.wav.
• The application only records while the record key is held down.
• The application only sends audio to Whisper when the translate key is tapped.
• The application does not transcribe audio that is less than 1 second long.
• The application does not handle errors from the Whisper API.