How I used AI to create an audiobook

Can you tell the difference?

If you read my last newsletter, you were probably wondering how I ever managed to finish my audiobook. That’s exactly what I’ll be going over in this post.

The emergence of AI Audio

AI audio tools have been around for a while but for the last few years, the only ones I was aware of were ones that helped with processing and editing. For example, in 2020 Descript came out with their software to allow for quick editing and automatic transcript creation. Descript did eventually launch a feature called Overdub that allowed you to create your own audio voice if you trained it with recording data (you had to read over an hour of audio they gave you). I tried this, but I found the generated audio to be very robotic.

However, in 2022 something exciting happened. With the boom of ChatGPT and the explosion of consumer AI tools, suddenly AI voice became a lot more realistic.

Creating an audiobook book with AI

The tool I got the most excited about was Eleven Labs, which lets you use your own voice with minimal effort! How it works is that you just need to upload an audio file of yourself talking for at least a few minutes (I had plenty of those from the hours I spent reading the first part of my book).

Surprisingly it’s pretty good. Pretty, pretty, pretty good! Is it 100%? No, but I would say it finally passed the uncanny valley phase and sounds good enough.

Finally finishing my audiobook

So, after getting good enough results from Eleven Labs, I decided to go ahead and finish my audiobook. I used AI to automatically record the remaining ~40% or so of the book that needed to get done, uploaded it onto ACX and it was good to go!

After I recorded all 10 chapters and submitted it to Amazon, I had to wait for them to do a manual review. I thought there might be a chance they reject it at that stage, but it went through smoothly!

You can check out my audiobook here or sign up to Audible if you’re curious.

Do I think the quality of AI audio now is at the same quality level it would have been if I recorded it myself professionally? No, I don’t. However, since this is just a side project, using AI did let me get it done and ship it, which is something I probably wouldn’t have had the time to do otherwise.

One of my favorite sayings that I live by is:

Finishing something at 80% quality is much better than trying to get to 100% and never finishing it.

Ahmad - Michael Scott

Stay tuned for next week’s newsletter, where I’ll get into what are high-level trends that I believe are changing in digital marketing (from AI to web3).

✌️ Until the next one,

Ahmad