Over the years, we’ve toyed with the idea of starting a podcast here at Zwift Insider. But it was simply too time-intensive of a project, so we opted to leave it to the experts.
Recently, though, I’ve been messing around with AI-based podcast generation tools. We all know AI is the biggest thing happening in tech right now – and it’s only getting bigger. What you may not know is that AI is revolutionizing the way podcasts are created, since AI can greatly reduce the time required to edit a podcast’s raw audio to remove unnecessary content and tweak audio settings for an optimal listening experience. (Simon used Adobe’s AI tools to edit the latest episodes of the Zwiftcast, in fact.)
I was interested in going one step further, though. Having already spent the time to write up detailed articles on so many Zwifty topics, I was looking for a way to auto-generate a full podcast. Just point the robot at key source articles, let it consume the content, and spit out a podcast with real (simulated) human voices.
Enter Google’s just-released NotebookLM, which does exactly that. And does it very well.
List to our first (let’s call it beta) episode, a 12-minute romp on the topic of Trainer Difficulty:
This podcast was created by pointing NotebookLM to the following posts:
- Using the “Trainer Difficulty” Setting in Zwift
- Using Zwift Trainer Difficulty to Simulate IRL Climb “Feel”
- What’s the Best Zwift Trainer Difficulty Setting for Zwifters?
- Zwift Trainer Difficulty on Descents: Did You Know?
- Trainer Difficulty on Zwift: How Much of a Bigger Granny Gear Does 75%, 50%, and 25% Provide?
- Wahoo Kickr Climb Zwift “Trainer Difficulty” Wheelbase Hack
As you’ll hear, NotebookLM has two voices (a male and female) who riff off of each other on the topic at hand. Having listened to this particular episode multiple times, I’m impressed by how they organize the content thoughtfully, summarize it, and even add little comments that I don’t remember writing in my original posts. Maybe that info comes from user comments on the articles? Or maybe the bots are more Zwift-smart than anyone realizes?
Your Feedback
I’m sharing this with the Zwift Insider audience because I want your honest feedback: do you believe a podcast with short episodes on specific Zwift features be useful to you, and/or the Zwift community at large?
Yes, no, maybe so… let me know in the comments below. You won’t hurt my feelings – after all, the episodes aren’t hard to create, and it’s not my voice you’re hearing!