A few days ago, I discovered NotebookLM and its audio overview feature and I was blown away!
I like to listen to podcasts… a lot. In the car, grocery shopping, jogging, doing chores, I always have a podcast on. It’s so efficient, you are learning while doing something else, even though I’m a basic podcast listener, listening at 1x speed (I have some friends listening to podcasts at 3x speed).
On a recent episode of the hardfork podcast, I learned about NotebookLM and its audio overview feature. It sounded super ineteresting, so I decided to give it a try. We have a lot of documentation for our RAC/M Identity solution (you can find it here) and I was curious to see if I could get NotebookLM to make it more useful. I simply uploaded the raw markdown of the documentation site (the site is built using Vitepress from raw markdown). Right away it built a FAQ and I could ask it questions about the docs that it could answer fairly well. I even asked about which standard features of Identity governance and administration solutions was missing from the documentation and it gave me interesting topics that are not well covered!
Then, I had it generate the audio overview and it built an entire podcast about our solution and it sounded good. I was so excited about this that I talked to everyone in the office about this new technology.
When I spoke with François-Xavier our VP of professional services, I insisted that he try it. So he did! He uploaded a document which describes the various activities that OKIOK offers in its offensive security (aka pentest) services.
Listen to this:
We put a lot of effort into describing accurately our activities in this document so that prospective clients can know everything that we do. But it is not very fun to read as its purpose is informational.
But in a few minutes, NotebookLM generated a podcast which is actually enjoyable to listen to. It’s very accurate (I’ve listened to it several times) and it doesn’t feel robotic. Listening to AI voices that sound robotic is not enjoyable and I can’t do it for more than a couple of minutes, but these voices sound very human.
I’ll also add the podcast about our IGA solution, but this one is a bit less accurate. First, they say R-A-C-M instead of the way that we say it “RACKEM”, there is also an analogy about keys that I don’t like too much. Also, its a bit off in the SaaS vs Managed services. But in general it does a fairly good job of taking hundreds of pages of documentation and boiling it down to a simple 12 minute podcast.
WARNING: This is not 100% accurate, it’s my first experiment.