Every writer knows that you can learn an incredible amount listening to your work read out loud. You hear the rhythm of the prose while also hearing what words or sentences are unnecessary.
All of my novels are available as audio books, and I can tell you that each time I’ve listened to my work read to me, I’ve found LOTS I wish I had changed or edited.
I have often longed to be able to listen to a book before publication. Narration is very pricey, however. Professionals get paid per word or by the hour and an author or publishing house can end up spending thousands of dollars for quality work (those professionals are TOTALLY worth it BTW*).
I’ve been very excited this week to try out a piece of AI that could be a game changer for me as a novelist. I’ve just finished a draft of a new mystery novel called Drowning Song. Eleven Labs, which I have written about before, offers text to voice generation for an entire manuscript!
Here is their AI narrator named Jessie:
Here is one of their many British women:
And, lastly, just for fun, we have this villainous man. It’s got a real Vincent Price, Tell-Tale Heart vibe that made me laugh.
A lot of the AI female narrators sound incredibly young, which, if you have been following this newsletter, is no surprise. But the idea of listening to ten hours of vocal fry makes me shudder.
I chose the first one, Jessie, despite it being a little upbeat for a mystery. You’re able to dictate if you want more or less voice variation, the equivalent of a director telling an actor “Give it a little more,” or “Take it down a notch.”
You could also go through the entire manuscript and dictate that different characters will have different voices. My book is over 100,000 words, so there is no way I would do this time-consuming task.
I did have some trouble downloading the final product and searched everywhere for a help line. There was none. You could interact with an AI chat bot about the problem. So AI was going to help me find the solution to AI? It’s like those “check all the boxes that contain stairs” to prove you’re human. A robot is testing me to see if I’m a robot. It’s one big AI ouroboros.
The chat AI bot was unable to answer my question, but after an hour of fiddling around with the software, I figured it out. I have my finished narrated book!
I will spend the next few days doing a puzzle as I listen to my book, pen ready to make notes on what I want to edit.
I confess I had some concern about giving AI my entire book. Is it now feeding its algorithm? Will it be shared with other AI? I copyrighted the entire thing this morning, just in case.
*I have friends who work in the book narration business and I just want to add that I would never choose these bots for a final version of a book. The bots are fine for a draft but they lack the intuition of an actor - the instinct of when to sound sad, angry, remorseful. These things are implied by the writing, but not in an overt way a bot could understand. People are still vastly superior.
NAILED IT
I was trying to get an image of a robot being chased by a mob with torches, classic Frankenstein stuff. AI had trouble. Check out the scale of the torches!
I tried to fix it, and instead it gave the torch to the robot, as if the mob wants to kill an Olympic runner.
I get a torch. You get a torch. Everyone gets a torch!
In this last one, the mob has made way for the robot and seems happy to light her path. Also, the woman on the left appears to have a robot head. Perhaps she is the long lost result of an tumultuous affair the robot had with a human?
Great tip!!!
That's absoluely unbelievable how good those VOs are already. I'm sad that it will eventually be able to do it well enough to put us actors out of work. But with the progress that's been made in such a short period of time, it feels like an inevitability.