How Google’s NotebookLM Brought My Sci-Fi Series to Life with AI-Generated Conversations 

I’ve been harnessing a variety of so-called “AI” tools as force multipliers in both my business ventures and creative projects. These tools have become indispensable as research assistants and writing companions, helping me push the boundaries of what’s possible. Recently, Google released an experimental tool called NotebookLM, and I was eager to give it a spin. As of this writing, the user interface offers a limited set of options, which is pretty much what you’d expect from a brand-new tool still finding its footing.

NotebookLM operates as a Retrieval-Augmented Generation (RAG)-locked model. Google assures us that it only uses the content you upload and doesn’t incorporate it into the model’s training data. This approach is reportedly designed to reduce hallucinations and errors that are common in other AI tools like Claude or ChatGPT. By focusing solely on your provided content, the model aims to deliver more accurate and relevant outputs.

One of the standout features is its ability to create an audio “deep dive” conversation using two very realistic AI-generated voices, all based on the material you upload. It plays out like a high-quality podcast or radio discussion centered around your content. It’s not just a monotone reading of text; it’s an engaging dialogue that explores various facets of the material.

 So, naturally, I couldn’t resist putting it to the test. I uploaded the first four books of my Accipiter War book series—a hefty 500,000 words—into the system. Admittedly, it took a while for the tool to process that much data, but the wait was worth it. What it produced honestly made my eyes widen in surprise. The voices are, well, exceptionally good. Shockingly good, to be honest. If I didn’t know any better, I’d think I was listening to two seasoned broadcasters discussing the series.

 The audio turned out to be a 15-minute deep dive—a back-and-forth conversation that zeroed in on some of the major characters and delved into themes, plot developments, character arcs, and more. The analysis didn’t shy away from tackling the moral, ethical, and philosophical explorations woven throughout the books. In fact, it was much deeper than I expected from an AI-generated discussion. There were a few things it didn’t quite grasp, such as the nature of the McKendree Cylinder world—an understandable oversight given the complexity of that concept—but overall, the depth and accuracy were impressive.

 Interestingly, the audio leaned toward being a bit of a fluff piece, offering an overwhelmingly positive review of the story. Not that I’m complaining; it didn’t exactly hurt my sense of pride in the years of work that Blake and I have poured into the series. Hearing our work discussed in such glowing terms, even by AI voices, was both surreal and gratifying.

 One thing that did catch me off guard, though, is that there doesn’t seem to be a transcript option. For a tool that processes and analyzes text, the absence of a written record feels like a missed opportunity. It would be great to have a transcript for accessibility reasons and for those who prefer reading over listening.

 Overall, my experience with NotebookLM has been fascinating. While it’s still an experimental tool with room for growth, the potential it demonstrates is enormous. I’m excited to see how it evolves and what new features Google will introduce in the future. If this is where AI-assisted content generation is headed, we’re in for some interesting times.