Among Google’s projects in the field of AI, there is a service called NotebookLM which has recently made headlines. As we mentioned in our article, NotebookLM is a note app or search tool that allows you to provide sources to the AI (such as a PDF document, the link to an article, a YouTube video, etc.). ), then ask questions whose answers will be based on the content of these sources. But if NotebookLM has gone viral, it’s mainly because this new service also includes a feature that automatically generates a podcast from the documents you provide to artificial intelligence.
And recently, according to the TechCrunch site, the Meta company has discreetly published similar technology on the platform GitHub. As a reminder, Meta is today one of the main competitors of Google and OpenAI in the field of generative AI. And the tool whose code was published on GitHub relies on the company’s Llama models, while NotebookLM relies on Gemini.
Called NotebookLlama, this new podcast generator is open source, but, for the moment, it is relatively simple. In essence, NotebookLlama is currently capable of generating podcasts from PDF files which will be analyzed by Meta’s AI, which generates a podcast script. Meta templates are also used to make the script more dramatic. And finally, other technologies are used to generate the audio file from this script.
Wow! Meta dropped an open NotebookLM recipe: NotebookLlama 🔥
It uses L3.2 1B/ 3B for pre-processing the PDF, L3.1 70B for Transcript creation, L3.1 8B for re-writes and Parler TTS for Text to Speech ⚡
Step 1: Pre-process PDF: Use Llama-3.2-1B-Instruct to pre-process the PDF… pic.twitter.com/L7hb5GsMtl
— Vaibhav (VB) Srivastav (@reach_vb) October 27, 2024
A tool to improve
For now, however, NotebookLlama still has flaws. Indeed, according to TechCrunch, the synthetic voice used is not natural enough. But on the GitHub page for this project, it is stated that it is possible to improve the quality of the podcasts generated by using other text-to-speech tools.
Otherwise, it is not certain that, like Google, Meta will launch a real product based on NotebookLlama. However, this project allows Mark Zuckerberg’s company to demonstrate that, like Gemini, its Llama models can also be used to generate podcasts.
- Recently, Google’s NotebookLM service, which automatically generates podcasts from the documents provided to it, has created a buzz on the web
- Meta has just released a similar tool based on its Llama templates called NotebookLlama
- NotebookLlama is open source, but, at the moment, it could still be improved
