Audiobooks are big these days. But what if you’re not in audio, either because your book didn’t sell or you’re self-published? Are you out of the loop?

Well, not necessarily thanks to AI which is popping up everywhere right now. Yes, you can use AI to turn your work into an audiobook. But is that a good idea? Is it really worth it?

First off let me state my bias. I’ve been fortunate enough to have my work narrated by some of the best in the business over the last twenty years. Performers like Richard Armitage and Alan Cumming, and lately Owen Findlay. All trained actors, as familiar with the stage as they are on screen, and that shows in their work.

Can you replace people like this with AI? Not fully, in my opinion. Narrators don’t just read a book. They interpret it and bring it to life, often in ways I never saw as the author who created it. I don’t think there’s an algorithm around that’s capable of that. But the technology is getting pretty close. And you may feel that close enough is good enough.

It’s your book and your decision. I’m not of the opinion that anyone who touches AI has handed their soul over to the devil. The truth is AI is becoming so ubiquitous you’re probably using it every day right now and not even noticing.

So your decision… and here are some of the things to consider.

The options

If you want to produce your own audiobook, there are three fairly different routes open to you right now. The cheapest and most obvious is Amazon’s own Virtual Voice, built into KDP — pick from around 80 stock voices, adjust pacing and pronunciation, and it’s live on Amazon, Audible and Alexa within minutes. But this is very, very basic and, to me, very obviously artificial. Your control of what’s being done is limited. You’re not directing a performance, just selecting and tweaking one.

Further up the scale sits ElevenLabs, which has spent 2026 pushing hard into audiobooks specifically. There’s a free tier that auto-converts your ebook for their own ElevenReader app, a self-directed studio, typically costing between $100 and $200 a book. You can cast narrators, clone a voice, add music and sound design, and there’s a fully managed production service for professional publishers that’s going to be beyond the budget of most individuals.

Then there’s the non AI option that hasn’t gone away: paying a human narrator through ACX, still the default route onto Audible.

The barrier that actually matters

Here’s the big gotcha you have to appreciate: ACX’s rules say submitted audiobooks must be narrated by a human unless otherwise authorised, and list unauthorised text-to-speech or AI recordings as prohibited. Use AI, and you’re immediately locking yourself out of the biggest and best-known audiobook platform around. Audible still holds somewhere around 60% of the US market, even as Spotify eats into that share. So the principal AI-narration barrier isn’t technical, it’s commercial. Doesn’t matter how well AI has handled the narration, your largest potential distributor won’t touch it with a barge pole, unless you go through Amazon’s own more rudimentary tools instead.

Spotify, Apple Books, Google Play Books and Kobo all accept AI narration provided it’s disclosed. But is that really enough? And there’s another complication we’ll come to shortly… the return may well stink.

What if you want it to sound like you?

I’ve tried audio narration. Trust me, it’s hard, time-consuming and difficult to get right both artistically and technically. But there are now ways with ElevenLabs you can ‘narrate’ in your own voice with AI. Cloning your own voice comes in two grades, and the gap between them is bigger than the marketing suggests. Instant cloning needs a minute or two of decent audio and gives you something usable within minutes — but it tends to flatten emotional range and natural rhythm, which matters over the length of a novel. In other words, it’s probably not good enough.

The proper version, Professional Voice Cloning, is a different undertaking entirely. You’ll need to record up to three hours of clean, studio-quality recording, at 44.1kHz/24-bit minimum, in a quiet or treated room, ideally in one continuous take rather than stitched clips. You also need to record in the tone you actually want for narration, since whatever style you feed it is the style you get back. Processing then takes several more hours on top. In short: to get an AI clone good enough for a full audiobook, you end up doing something that looks a lot like recording an audiobook.

And the money?

This is the part you really need to think about. What are your chances of making much of a return? Audiobook income splits into two completely different mechanisms, and only one of them pays well.

Sell a copy outright and you get a proper cut of the price — 50% on Spotify’s direct sales, 60% through ElevenLabs’ own store, comparable rates via ACX, Google Play or Kobo. That’s a real royalty worth having.

But most listening on Spotify happens through a subscriber’s monthly allowance, not a purchase, and that’s paid from a shared pool split pro-rata across everyone’s listening that month — the same mechanism, and the same dynamics, that leave musicians earning fractions of a cent per stream. If you want to know how that works out in reality, ask a musician about Spotify and stand well back before they answer.

ElevenLabs’ own streaming payout, through their ElevenReader app, is quoted at $0.20 per hour listened — and even that headline figure doesn’t sit comfortably with an earlier, much higher number the company floated during its beta. Nobody’s publishing a rate that makes streaming income easy to plan around.

Does that make AI narration pointless? Only you can decide. I know I’d get killed by my listeners if my work went to AI because they’re used to top notch narration from my long-term audio publisher WF Howes. But that’s how it stands for me. Others may feel it’s a real way to get a backlist title into audio at a fraction of the traditional cost, and reach listeners a print or ebook edition never will.

My feeling… don’t get sucked in by the hype, and do expect some blowback from the ‘all AI is bad’ people out there. From my point of view as an established author, any arrangement that would keep me off Audible is a non-starter. Others will doubtless feel differently, as is their right.