/paper_to_podcast

A very quick and intuitive project that transforms any research paper into an engaging discussion between 3 persons to make you listen and have intuitive thoughts about the paper instead of using a paper reader. This is the perfect project for podcast lovers like me.

Primary LanguagePythonApache License 2.0Apache-2.0

Paper-to-Podcast 🎤

1000004188

Paper-to-Podcast is a tool that transforms academic research papers into an engaging and conversational podcast format. With this project, listeners can absorb the content of a research paper in a lively discussion involving three distinct personas—perfect for those who prefer listening over reading, especially during commutes or travel.

Project Overview

Objective

This app simulates a three-person discussion around the content of a research paper, making complex information more accessible and enjoyable to absorb. Instead of merely reading aloud, it converts papers into conversations that are engaging and intuitive, providing valuable insights and critical thinking.

Personas

  • Host: Guides the conversation, introducing each section and explaining the main points in an engaging and warm tone.
  • Learner: Asks intuitive questions and brings curiosity to the discussion, helping listeners grasp core concepts.
  • Expert: Provides in-depth knowledge and additional details, enhancing the discussion with profound insights.

This structure fosters an interactive listening experience, helping users better understand the paper in a way that feels natural and human.

Code Structure and Key Components

  • Planning Chain: Starts by creating a detailed plan for each section of the paper. Planning helps the model stay on track, reducing the chances of hallucinations or redundancy.
  • Discussion Chain: Uses a retrieval-augmented generation model to expand on each section. This ensures the script stays true to the source content while generating meaningful dialogue.
  • Enhancement Chain: Finalizes the script by removing redundancies, refining transitions, and ensuring a smooth flow.
  • Text-to-Speech: The generated script is then converted into audio using the OpenAI API, producing realistic voices for each persona. image

Cost Efficiency

The app is cost-effective, utilizing OpenAI's API. For example, generating a 9-minute podcast from a 19-page research paper costs approximately $0.16.

Usage Instructions

Prerequisites

  1. Clone this repository: 
    git clone https://github.com/Azzedde/paper_to_podcast.git
  2. Move into the project directory: 
    cd paper_to_podcast
  3. Ensure you have a valid OpenAI API key stored in your .env file.

Running the App

  1. Place a research paper in PDF format in the project directory.
  2. Run the script from the terminal, providing the path to your PDF file as an argument: 
    python paper_to_podcast.py path/to/your/research_paper.pdf

Sample Podcasts

You can find examples of podcasts generated using this pipeline in the ./sample_podcasts directory.

Roadmap

  • Optimization: Currently, the process takes times. Further optimization is planned to reduce runtime.
  • Local LLMs and TTS: Exploring alternatives to OpenAI’s API for a completely free, local implementation using Ollama and open-source TTS models.

Contributing

If you’d like to contribute, there is an open issue for optimizing the podcast generation time. Feel free to explore or create new issues to enhance the app!