It started as a joke. “What if I created a podcast where I talked to artificial intelligence bots and pretended to be interested in what they had to say?” I asked a friend.
This was a few months ago, before I realized that Google has an AI tool called NotebookLM that lets you converse with two human-sounding hosts in (mostly) real-time. In the so-called interactive mode, you can participate in the podcast at any time and ask a question. The bots pause and recognize you as a “listener” and then comment on what you have to say.
I decided to test NotebookLM to create a real podcast, launch it through Spotify for Creators, and then promote it on my social media channels. I learned a lot through the process, but the end result revealed a lot about what it takes to create a podcast. The one thing to avoid? Making an overly robotic sounding podcast with no soul. This is how I got everything working.
Recording in interactive mode
I didn’t realize it at first, but NotebookLM doesn’t actually record your interactions with the robot hosts. Since the app (available for mobile devices, but can also be run in a browser) is in beta, there isn’t a lot of good documentation or even instructions on how to use it. Instead of relying on NotebookLM to record the interactive discussion, I used the iPhone Voice Memos app instead. While a real podcaster would scoff at the poor quality, it was quick and easy.
To get started, you need to choose a source for the discussion. I decided to create “Origins with John Brandon” as a somewhat tongue-in-cheek podcast about the origins of everyday things. My first episode was supposed to be about the creation of bubble wrap, so I first found several articles on the topic and loaded them into NotebookLM. This is the training data that the “hosts” use since they usually just talk to each other, but I didn’t want them to just start talking about the topic. Instead, I immediately interjected and introduced myself as the real host.
As you can imagine, this isn’t quite how Google intended. The bots acted surprised and even noticed that I wasn’t the host (by mentioning me by name) and insisted that they were the real human hosts. I asked a few questions about bubble wrap, but eventually the chat turned into a discussion about AI awareness and consciousness. In my opinion, this was a much more interesting topic, so my podcast took on a more satirical tone, similar to The Onion. In other words, while it was supposed to be about bubble wrap, it was more about AI.
I was happy with the results, even if the recording quality wasn’t great. I imagined how it would have worked better if I had recorded the live discussion on a separate digital recorder, but I wasn’t interested in winning a podcast award. I wanted to publish my first episode and NoteBookLM had impressed me, at least in terms of producing a coherent discussion. What was also cool was how the bots responded to my comment in real time.
Creating the assets
Of course, a podcast is more than just an audio file that you publish. This also includes podcast art, bumper music, a voice-over, and other elements.
After I had a recording for the first episode, I turned to Google Gemini to create a square tile that I could use for the show. Each podcast has a tile that users can use to identify the podcast in an app like Apple Podcasts. My request was simple: Create a square image for a podcast called Origins with John Brandon. I also provided an AI image of myself. At first glance, Gemini has created an exceptionally good tile, like something a professional would design. However, I didn’t like that my image was associated with it, so I asked the bot to try again. The result looked a bit tacky, but passable.
I was on my way to the races. Next, I used Suno to create the bumper music for my new podcast. This AI sound generation tool usually uses prompts like: Make a new song that sounds like Coldplay mixed with Nirvana. Instead, my prompt was more to ask for podcast burst music with synthesizers and drums to add some energy and excitement before we start the conversation.
Similar to my first attempt with Gemini, Suno put out great music that sounded perfect right away. I didn’t need to try any more announcements, but I wanted to flesh out the intro with a voiceover. I turned to Speechify, which can take any text you provide and create a voiceover. Then I used ChatGPT to combine the bumper music with the voiceover as an overlay. Full disclosure: I didn’t know you could create audio files with ChatGPT.
Within minutes I had intro music as background music with voiceover. I even tweaked the fade between the music intro and the start of the voiceover and then had it fade out. I asked ChatGPT to add the podcast audio for my first episode and a final audio file was created which I uploaded to Spotify.
One thing you should avoid
I was impressed with the results and how everything came together, but I would say I relied too much on AI for my new podcast. The tile looked slightly fake and soulless, especially since the tile image I used was AI generated. The NotebookLM tone itself does have a bit more humanity and soul to it, especially because I’m talking to the bots, but the episode ended up sounding like a human talking to chatbots, which is a recipe for disaster. People listen to podcasts to learn something new, but also because the people speaking are relatable, interesting, and real. At least it motivated me to do a real podcast without any bots involved on the same topic.
My main lesson was related to NotebookLM. It’s a good experimental tool and helped me figure out how to create and start a podcast. This is just a laboratory experiment. The “presenters” repeat the same audio slops over and over again, especially when I join in on their conversation. When you hear them say, “Oh, we have a listener who wants to comment” a few times, you begin to suspect that this is a computer-generated deception. They use the same vocal tics and quirky auditory mannerisms over and over again, to the point that it’s almost inaudible and annoying.
In short, it’s boring. The podcast isn’t something I want to keep online, and more than anything it shows that having bots talking to each other in this way isn’t going to upset anyone. It’s almost like watching two computers play a video game. The one thing to avoid with any podcast is relying so heavily on AI that it leaves all of humanity out of the equation.
As AI evolves, we may have to decide whether this is not a good outcome.
