open-notebooklm: Turn PDFs Into Podcast Audio

Overview

Open NotebookLM is a Gradio application that converts PDF content into a conversational podcast script and audio file. It uses an LLM to generate dialogue and text-to-speech models to produce an MP3.

It is intended for people who prefer listening to document material or want a more conversational way to review a PDF. The project is inspired by NotebookLM and is built with open-source language and speech models, while relying on the Fireworks API for LLM access.

Key Features

• Accepts a PDF as the source for a generated podcast.
• Creates an informative, entertaining dialogue from document content.
• Produces an MP3 file from the generated dialogue.
• Provides a Gradio interface for uploading documents and starting conversion.
• Uses Llama 3.3 70B via Fireworks AI for dialogue generation.
• Uses MeloTTS and Bark for speech generation.
• Described as available in 13 languages in the project’s Hugging Face Space metadata.

Use Cases

• Students can listen to a conversational review of course PDFs while commuting or doing other tasks.
• Researchers can create an audio overview of a paper before deciding which sections to examine closely.
• Professionals can turn reports or briefing documents into audio for review away from a screen.
• Readers who find dialogue easier to follow than dense prose can use it to engage with PDF material in audio form.

Getting Started

git clone https://github.com/gabrielchua/open-notebooklm.git
cd open-notebooklm
pip install -r requirements.txt

Set the FIREWORKS_API_KEY environment variable, then run python app.py. See the README for setup details and usage instructions. A hosted interface is available on Hugging Face Spaces.

Alternatives

• podcastfy: Podcastfy turns PDFs and other sources into multilingual conversational audio, while open-notebooklm focuses on generating a podcast dialogue and MP3 from a PDF.

Considerations

• Running the application requires a Fireworks API key, so it depends on an external LLM service rather than being fully local.
• PDF conversion involves multiple generation steps, and the resulting dialogue and audio may not preserve every detail or nuance of the source document.
• The repository’s latest recorded push was in December 2024, so check the repository for current compatibility and maintenance information.