Can you use an AI voice generator to create an entire podcast episode?

With all of the excitement around artificial intelligence, many podcasters are asking:

"Can you use an AI voice generator to create a podcast episode?"

I wanted to put that question to the test, so I generated audio with a few popular AI voice tools.

The first experiment I ran was a bot that a local college kid put together. I prompted it to "create a podcast episode for a podcast about extreme rollerblading. Talk about the rollerblading scene in Stony Plain, Alberta," (which is my hometown). 

Using only a 30-second clip of my voice he found on YouTube, he was able to generate AI audio content that sounds very much like me. What's most surprising to me is how the AI synthesized not only my voice but also my cadence, my pauses, and my tone.

However, I'm still unsure that a longer episode using this AI voice would be entertaining. That's really the question podcasters (and aspiring AI entrepreneurs) are asking: can you generate compelling long-form AI audio content?

The first AI podcast experiment I saw was from the folks at podcast.ai. Using AI voices, they generated an interview between Joe Rogan and Steve Jobs. 

To my ear, it sounds stilted; while it's a good demo of how far this AI voice generation has come, it's still not at the place where it makes for compelling entertainment.

Another interesting experiment was from the Mind Meets Machine podcast. In this case, there's a human co-host (Rob) and an AI co-host (Ruby). 

What these AI podcast experiments are missing is, ironically, human connection. There's something about hearing somebody in your earphones revealing deeply human moments that's hard to replicate with AI.

Another popular clip was an AI version of Bill Gates and Socrates.

Again, most of these AI clips don't sound natural; they're not quite there yet when it comes to producing something that I think could be entertaining, especially in a way that would have a podcast listener coming back episode after episode.

As a part of this experiment, I also tried some of the popular AI voice generators on the market: Genny and Play.ht. Here's a sample of each:

Can you use an AI voice generator to generate a compelling podcast?

Currently, the verdict is clear: AI voice generation isn't going to displace human hosts. Why? Because when we tune into podcasts, we seek more than just content. We're not looking for mundane, run-of-the-mill audio. We want authentic, gripping stories that resonate with our emotions, offering human connection, drama, and depth.

While AI voice generators have made strides, they still lack that genuine human touch. The nuances, the warmth, the imperfections – they're all absent.