audio-summary

audio-summary is a command-line tool designed to create meeting minutes from an audio recording, leveraging OpenAI Whisper for transcription and Google Gemini for summarization. Currently, only an online version is available, with plans for an offline version utilizing Ollama and HuggingFace under development.
Demo

Dependencies

Ensure you have ffmpeg installed:

# Mac
brew install ffmpeg

# Windows 
# 至ffmpeg 官網下載安裝: https://ffmpeg.org 

Installation

# Git
git clone https://github.com/thisishugow/audio-summary.git
# pip
pip install ./audio_summary.whl

Configuration

Create .env file and setup OpenAI API Key and Gemini API Key

OPENAI_API_KEY=your_OpenAI_API_key
GOOGLE_API_KEY=your_Google_API_key

Usage

• Streamlit UI

  python -m audio_summary.server

• Use command line

  python -m audio_summary -f meeting-recording.wav -s true

  Options:

  • -h, --help: Show help message and exit.
  • -f FILE, --file FILE: Specify the path of the audio file.
  • -oOUTPUT, --output OUTPUT: Specify the path of the output transcription.
  • -s SUMMARIZE, --summarize SUMMARIZE: Specify whether to use Gemini for summarization (true/false). Default=true.
  • --summarize-by API, : Specify the summarization API to use. Choices: openai, gemini. Default=openai.
  • --lang LANG let AI response in ["original", "en", "zh-tw"]. Default="original" Then you will see the full transcription and the meeting minutes.

The tool supports summarization using either Google Gemini or OpenAI's models. You can select the preferred provider using the --summarize-by argument in the command line or via the UI.