After the fuss last month that greeted the release of Google’s experimental AI podcast generator NotebookLM, it is no great surprise to see fellow tech giant Meta release its own alternative product.
Like its forebearer, NotebookLlama from Meta enables users to convert a text file – such as a PDF or Word file – into a fully formed, AI generated podcast with multiple ‘hosts’ carrying out a discussion on its contents.
Although early reports suggest that the end result produced by NotebookLlama doesn’t sound as polished as NotebookLM’s output, the major difference between the two is that Meta has made its version open-source.
Creating Drama With NotebookLlama
Meta shared files relating to NotebookLlama on GitHub, together with a full tutorial and ‘PDF to Podcast’ workflow.
The page says that the tool can be used by anybody as it “assumes zero knowledge of LLMs, prompting and audio models”. All the instructions and knowhow are contained in the notebook.
This just in! View
the top business tech deals for 2024 👨💻
It then gives an example of how a podcast can be created in four steps, together with the specific prompts to use to ‘pre-process’ (i.e. clean up) your PDF, produce an initial transcript, dramatize it and then, ultimately, generate the conversational podcast.
Limitations and Improvements of NotebookLlama
In addition to giving more detailed steps on using NotebookLlama, the GitHub page also lists next steps and improvements for the tool.
These include better prompting, support for ingesting information from other sources (e.g. websites, YouTube videos, etc) and different podcast formats.
Perhaps most crucially in order to match the splash made by NotebookLM, it notes that there is currently a limitation on how natural the speech model sounds and acknowledges that this will probably be improved with “a better pipeline”.
It was the sense of natural, human-like interactions that most wowed the tech community when NotebookLM first surfaced. Meta’s competing tool is a little way behind that for the time being, with commentators noting that the “voices have a robotic quality” and “the AI speakers unintentionally talk over each other“.
Open Source Provides Path Forward
NotebookLlama continues Meta’s commitment to open sourcing its AI technology – it did exactly the same thing when it launched its Llama 2 large language model last year.
Founder and CEO Mark Zuckerberg explained in a blog post in July why he believes open-source is the “path forward” for research and development in the area, citing the precedent of Linux’s role in the development in cloud computing and operating systems.
“We’re taking the next steps towards open source AI becoming the industry standard.” – Mark Zuckerberg
In it, he argues that open source AI is good for developers, Meta and the world. “AI has more potential than any other modern technology to increase human productivity, creativity, and quality of life,” he writes.
“Open source will ensure that more people around the world have access to the benefits and opportunities of AI, that power isn’t concentrated in the hands of a small number of companies.”
