Meta NotebookLama AI podcast generator released as ‘open source tool’ to compete with Google’s NotebookLama

Meta on Sunday released a new open-source artificial intelligence (AI) tool that will compete with Google NotebookLM. The tool, called NotebookLama, is an AI-powered podcast generator where users can upload a PDF file and the tool will turn it into an audio podcast with two AI characters. The tool uses three different Llama 3.1 AI models to complete the entire process. Like Google’s tool, NotebookLlama’s podcast also follows a back-and-forth conversation between two AI hosts in a free-flowing manner.

The Meta NotebookLlama AI tool uses three large language models to generate audio podcasts from blocks of text. Currently, the tool only accepts PDF files as input, so users will have to convert whatever text format they have into PDF.

Meta NotebookLama Workflow
Photo Credit: Meta

NotebookLlama uses the Llama 3.2 1B instruction model to first pre-process the PDF file and save it to a ‘.txt’ file. The Llama 3.1 70b instruction model is then used to write the podcast transcript using the source dataset. The transcription is then dramatized using a rewriter that uses the Llama 3.1 8b instruction model. Finally, a custom tool is used to add transcription to the text-to-speech workflow. For this Meta Parler is using TTS tool. Interested individuals can access all the models needed to generate a podcast from the GitHub listing here.

However, the AI ​​models mentioned above are merely recommendations from the developers. Users may prefer to use smaller models for each step, however, results may vary. Meta highlighted that to run the AI ​​system in the recommended setup, users will need a GPU with around 140GB of total memory.

An X (formerly known as Twitter) user posted a sample of a generated podcast. Based on this, it appears that the audio quality is not as good as Google NotebookLM, and sounds tinny and robotic. Additionally, there are instances where parts of the audio are skipped and the AI ​​hosts end up speaking over each other.

Meta acknowledges some issues and plans to improve them in the next version of the AI ​​product. The company highlighted, “The TTS model is at the limit of how natural it will feel. “It could probably be improved with better plumbing and help from someone more knowledgeable.”

The tech giant also plans to use two different LLMs to write scripts, where each model will debate the other to make the podcast sound more conversational. It is also part of the developers’ future pipeline. Additionally, the company is also testing the Llama 405B AI model to enhance support for more input and output formats while writing transcripts.