Key Takeaways
- Similar to Google’s NotebookLM, Meta introduces NotebookLMa, an open-source podcast generator.
- With the help of two AI voices, NotebookLlama transforms PDFs into conversational audio.
- The gadget sounds more robotic than human, according to early reviews.
NotebookLlama, an open-source AI podcast generator developed by Meta, converts PDF pages into audio dialogues.
The public can download the solution, which was created using Meta’s Llama model, according to files posted on GitHub.
Meta’s NotebookLlama allows users to convert documents into conversational audio with two AI voices, much like Google’s NotebookLM, which assists users with a one-click podcast production option.
NotebookLlama uses Meta’s Llama 3.1 models in a sequence of stages to handle user-uploaded PDF files. The Llama 3.2 1B model first converts the PDF to text format. A smaller Llama 3.1 8B model adds a conversational tone, and the Llama 3.1 70B model then uses the text to create a podcast script. Lastly, a dynamic dialogue between the AI characters is created by turning the writing into audio using Meta’s Parler TTS tool.
NotebookLM Outpaces NotebookLlama
In early feedback, many expressed their enthusiasm and commended the tool’s potential, but others pointed out a number of drawbacks.
On X, users complained that the audio results were not as seamless as those from NotebookLM, with sporadic overlaps where the AI speakers inadvertently spoke over one another.
Meta has acknowledged these issues and promises to enhance the dialogue flow and sound quality in its NotebookLlama roadmap. Among the planned enhancements is the use of distinct language models for every AI character to more accurately replicate back-and-forth conversations while maintaining a human-like audio quality.
As a direct rival of Google’s NotebookLM, NotebookLlama fits in with Meta’s larger open-source approach in the AI space, where the Facebook parent company has made a number of other open-source AI models available, including DINOv2 and ReAgent.
