GPT Summarisation App

The GPT Summarization App is a powerful tool designed to streamline the process of extracting and summarising content from YouTube videos. By leveraging the capabilities of OpenAI's Whisper model for accurate transcription and the GPT-3.5 API for concise summarisation, this app offers a seamless workflow for users to input a YouTube URL and receive a summarised text of the video's audio content.

Features

• YouTube Audio Download: Automatically downloads the audio track of any YouTube video provided via URL.
• Audio Transcription: Utilises the Whisper model for state-of-the-art audio transcription, ensuring high accuracy in converting speech to text.
• Content Summarisation: Leverages the GPT-3.5 API to produce clear, concise summaries of the transcribed text, making it easy to capture the essence of the video content.

Getting Started on MacOS

1. Clone this repo

2. Create a virtual environment with venv

   python -m venv .venv && source .venv/bin/activate && python -m pip install --upgrade pip

3. Install dependencies

   pip install -r requirements.txt

4. Install homebrew

5. Install ffmpeg with brew install ffmpeg

6. Download the spacy model:

   python -m spacy download en_core_web_sm

7. Install insanely-fast-whisper for transcriptions:

   pipx install insanely-fast-whisper --force --pip-args="--ignore-requires-python"

8. Create .env in root directory and add your OpenAI API key:

   OPENAI_API_KEY='add your OpenAI API key here' # https://platform.openai.com/account/api-keys

Usage

• Run the app from the command line, passing the YouTube URL as an argument. To download and transcribe a Youtube video, run this command:

  python -m gpt_summariser.download_and_transcribe <youtube_url>

• To download and summarise a Youtube video, run this command:

  python -m gpt_summariser.download_and_summarise <youtube_url> <title>

Repo structure

.
├── LICENSE
├── README.md
├── gpt_summariser
│   ├── __init__.py
│   ├── download_and_summarise.py
│   ├── download_and_transcribe.py
│   ├── download_yt_audio.py
│   ├── summarise_transcript.py
│   ├── transcribe_audio.py
│   └── utils.py
├── outputs
│   ├── audio
│   ├── summaries
│   └── transcripts
└── requirements.txt

TODO

• Download the YouTube video as a wav audio file
• Transcribe audio file using OpenAI's Whisper model to txt or vtt formats
• Summarise the transcript using free models from HuggingFace
• Transcribe audio file using insanely-fast-whisper and distil-whisper for faster transcriptions
• Add documentation
• Add tests
• Use free open-source models for summarisation instead of GPT3.5
• Add speaker diarisation support
• If the YouTube URL is in the youtu.be format, then convert it to the /watch?v= format
• Output transcripts and summaries into logseq markdown format
• Add front-end UI
• Add support for pdf and epub formats
• Batch downloads and summarisations
• Dockerise app

Contributing

Contributions are welcome! If you'd like to improve the GPT Summarisation App, please fork the repository and submit a pull request with your proposed changes.

License

This project is licensed under the MIT License - see the LICENSE.md file for details.