https://huggingface.co/spaces/gabrielchua/open-notebooklm
This project is inspired by the NotebookLM tool, and implements it with open-source LLMs and text-to-speech models. This tool processes the content of a PDF, generates a natural dialogue suitable for an audio podcast, and outputs it as an MP3 file.
Open NotebookLM turns PDFs into dynamic podcast-style audio files using Llama 3.1 405B for natural language generation and MeloTTS for text-to-speech. Note that only the text of the PDF is processed and there is a 100 000 character limit.
Key features
- PDF to podcast: Upload any PDF, and it’s automatically transformed into an audio file.
- Interactive dialogue: The generated dialogue simulates a podcast conversation, enhancing listener engagement.
- User-friendly interface: Users can interact with the tool through a simple web interface, even without deep technical knowledge.
Open NotebookLM is impressive but lacks some of the advantages of NotebookLM, like integration with Google’s ecosystem, support for Google Slides, and web URLs. In addition, Google’s computational resources and proprietary AI models enable advanced features like fact-checking and study guide generation, which are currently beyond Open NotebookLM’s capabilities.
Open NotebookLM…while not as feature-rich as Google’s version, offers an important alternative for users who value data privacy and the flexibility of an open-source approach (Kumar, 2024).
Open NotebookLM allows users to test the tool via its Hugging Face page or install it locally from its GitHub repository.