The reliable narrator

Published January 9, 2023

A 3D render of a cute orange robot sitting at a desk waving, with a laptop sitting on the desk in front of it

The Guardian published an interesting post last week about audiobook narration and the impact artificial intelligence could have on it. The post talks to an author who was working with Apple on a secret project that allows AI voices to provide the narration on an audiobook - and evaluates whether it’s any good or not.

From sentence to sentence, the cadence of the narrator’s voice glides forward, then snags on an artificial syllable. It’s the aural equivalent of watching the gears of a machine rotate under a surface of what looks like human skin.

It’s no surprise to see it doesn’t completely convince as a replacement for the human voice, but AI speech has come a really long way in a few short years so it’s only going to get better from here. The difference on voice assistants like Siri now compared to just a few years ago is night and day. I can’t see a real problem with listening to an audiobook narrated by a synthesised voice, other than my feeling is it might get grating after a while, so you’d probably have to pause than you might otherwise.

It’ll reduce costs drastically. Hiring someone to sit and read a book for many hours doesn’t come cheap and the price of audiobooks reflects that. You can argue that it’s a fair price for the hours of work gone into it, but you can’t argue that it’s accessible to as many people as it could be. AI could help lower those barriers to get people into listening books, even if it’s not available on every title.

But cheaper AI narration is likely to be a good option for readers who care about cost and who “may not necessarily need the fully narrated drama experience”, she says. “A lot of people are becoming more used to listening to these voices.” And the quality of the AI speech may not matter as much, she adds, “if you are listening to the book at 1.5 speed, which I do when I walk”.

I have to admit you’d be losing something, though, from the best narrators out there. There’s something very special when an author reads their own work, particularly if they have a good voice or if they’ve had training themselves. And if you manage to get that magic combination of an actor turned narrator and a wonderful book, well, it’s something indescribable.

There won’t be any of that with AI narration, and of course there’s an argument that a reader without the understanding of what they’re actually trying to get across is going to miss out on impact and gravitas. But at least they’ll be consistent. It still makes me rage sometimes when you get weird pronunciations in audiobooks or something doesn’t quite work with how you thought about it in your own head.

I can’t see it taking off in the immediate future but this is an interesting area to watch develop, and another use for AI that I hadn’t considered before.

← Previous The best godmother in the world
Next → Plus a bit more