How to make an audiobook in 2026: ACX vs Virtual Voice vs human narration
The three routes to an audiobook — hire a narrator, do it yourself, or use Amazon’s AI Virtual Voice — compared on cost, quality, rights and distribution, with an honest look at where a human voice still wins.
There are three honest routes to an audiobook. Hire a professional narrator through Audible’s ACX — either paying a per-finished-hour fee (roughly USD 100–400+ an hour) or splitting royalties 50/50 on a seven-year exclusive deal. Narrate it yourself. Or generate AI narration free through Amazon KDP’s Virtual Voice. Which fits depends on your budget, your genre and how much the voice matters.
That is the decision in a sentence. Here is the version that helps you actually choose — including what changed in 2026, and where a human reader still earns the money.
What are the three ways to make an audiobook?
Strip away the marketing and every audiobook is made one of three ways. You pay a person to read it well, you read it yourself, or you let a machine read it. Each trades cost against quality and control, and the right answer is genuinely different for a literary memoir than for a how-to guide.
The audiobook is a separate edition built from the same manuscript that becomes your eBook and print book — so it sits naturally after those are done. If you are still sequencing your launch, our guide to whether to launch the eBook or print first sets the order, and the wider step-by-step guide to self-publishing a book shows where audio fits in the whole path.
How does hiring a narrator through ACX work?
ACX (Audiobook Creation Exchange) is Amazon’s marketplace connecting authors with narrators and producers, feeding Audible, Amazon and Apple Books. You post your book, audition narrators, and agree one of two payment models.
The first is pay-per-finished-hour (PFH): you pay a flat fee for each hour of finished audio, regardless of how long it takes to record. In 2026, indie and emerging narrators typically charge from around USD 100 PFH; the SAG-AFTRA and ACX union minimum sits at USD 250 PFH, and seasoned or premium voices run USD 400 and well beyond (ScribeCount). A 70,000-word novel runs roughly 8–9 finished hours, so a mid-range narrator is a four-figure outlay. You own the recording outright and can distribute it anywhere.
The second is royalty share: you pay nothing upfront and split audiobook royalties 50/50 with the producer for a seven-year initial term. The catch is that royalty-share titles must be exclusive to Audible/ACX — you cannot sell them on Spotify, Apple direct, or anywhere else — and both parties must agree to end the deal before the seven years are up (ACX). It lowers your cash risk and hands a chunk of upside, and a long lock-in, to someone else.
What changed with ACX royalties in 2026?
This is the part to get current on, because it shifted recently. From 26 May 2026 Audible moved to a new royalty model, open to all ACX creators, with the legacy model retiring by the end of the year — so you will have to choose. Headline rates rose: 50% on exclusive distribution (up from 40%) and 30% non-exclusive (up from 25%) (ACX).
The bigger change is how you are paid. Instead of a fixed percentage of each sale, the new model pays from “Member Value” — a listener’s monthly subscription, minus taxes and fees, shared out across the titles they actually engage with. Your earnings now track listening, not just purchases. Because these terms are volatile and US-centric, confirm the live figures in your ACX dashboard before you commit; treat any number here as a starting point, not a promise.
What is KDP Virtual Voice, and is it any good?
Virtual Voice is Amazon’s AI narration, generated from your Kindle eBook inside KDP. At the time of writing it is a beta open to selected US authors publishing in English, and it is striking on cost and speed: it is free to produce, you pick from around 80 synthetic voices (you can even set a different voice per chapter), and the audiobook can be live in hours rather than weeks (KDP).
The limits are real. Only live Kindle eBooks qualify — no paperbacks, pre-orders or public-domain titles — and the book must be under roughly 240,000 words (about 26 hours). You earn a 40% royalty on a-la-carte sales, with the list price set between USD 3.99 and 14.99 (KDP). Virtual Voice titles are clearly labelled as such in the store, so listeners know what they are buying. Because it is a US-only beta, availability and terms will change — check KDP for your market.
ACX, Virtual Voice and DIY compared
| Factor | ACX human narrator | KDP Virtual Voice (AI) | DIY (narrate yourself) |
|---|---|---|---|
| Upfront cost | USD 100–400+ per finished hour, or 0 on royalty share | Free (beta) | Gear only (mic, quiet room) |
| Time to publish | Weeks (audition, record, QC) | Hours | Weeks to months of your time |
| Voice quality | Highest — performance, character, emotion | Clear but flat; audibly synthetic | Variable; depends on your skill |
| Royalty | 50% exclusive / 30% non-exclusive (2026) | 40% a-la-carte | 50% / 30% via ACX, your call |
| Rights and lock-in | You own it (PFH); royalty share = 7-yr exclusive | Amazon ecosystem; clearly AI-labelled | Fully yours |
| Distribution | Audible, Amazon, Apple; wide if non-exclusive | Audible / Amazon | Anywhere you upload |
| Best for | Fiction, memoir, anything voice-led | Non-fiction, guides, low-stakes backlist | Authors who can genuinely perform |
Where does human narration still win?
On the things a listener actually feels. A skilled narrator does not read a book — they perform it: distinct voices for each character, comic timing, the held beat before a reveal, grief that catches in the throat. AI narration has improved fast and is genuinely serviceable for instructional non-fiction, where the job is clarity, not character. But it cannot act. It flattens dialogue, fumbles sarcasm, mispronounces the invented name in your fantasy series, and never once decides that this line should land softer than the last.
For fiction, memoir and anything carried by voice, that gap is the whole product. Audiobook listeners are intimate listeners — earbuds in, hours at a time — and they notice the difference between a person and a plausible imitation of one. The same reasoning runs through our work on the page: we believe creative craft is human, which is why our covers, illustration, writing and editing are made by people, never by AI. A narrator is the audio equivalent of that conviction. If your book lives or dies on its voice, generate the cheap version to test the waters by all means, but the edition you are proud of is read by a human.
How do you distribute an audiobook?
Through ACX, your audiobook reaches Audible, Amazon and Apple Books. Choose exclusive and you are locked to those for the higher 50% rate; choose non-exclusive (30%) and you are free to also sell through aggregators that push to Spotify, Kobo, libraries and beyond. Virtual Voice keeps you inside Amazon’s ecosystem. DIY recordings can go anywhere you can upload a file. The decision mirrors the eBook trade-off between reach and rate — weigh how much a wider catalogue is worth against the better per-sale cut.
So which route should you pick?
If your book is voice-led and you can afford it, hire a narrator on pay-per-finished-hour, keep the rights, and distribute widely. If cash is tight but you believe in the book, royalty share gets it made for free — accept the seven-year exclusivity as the price. If it is practical non-fiction or a backlist title you simply want available in audio, Virtual Voice is a sane, free way to be there. And narrate it yourself only if you can truly perform, not merely read.
Librum does not produce audiobooks — but the audiobook is only as good as the manuscript and the eBook beneath it, and that is exactly what we make: a clean, edited book and an EPUBCheck-validated reflowable ePUB 3 that becomes the source your narrator reads from. See what that includes on our services page, or browse our packages to get the written word right first.
Frequently asked questions
How much does it cost to make an audiobook in 2026?+
It ranges from free to several thousand. Amazon’s KDP Virtual Voice generates AI narration at no cost during its beta. A human narrator through ACX charges roughly 100 to 400-plus US dollars per finished hour, so a typical novel runs into four figures, or nothing upfront if you choose royalty share and split earnings instead.
What is the difference between ACX royalty share and pay-per-finished-hour?+
Pay-per-finished-hour means you pay the narrator a flat fee for each hour of finished audio and own the recording outright, free to sell it anywhere. Royalty share means no upfront fee but a 50/50 split of audiobook royalties for a seven-year term, and the title must stay exclusive to Audible and ACX throughout.
Is KDP Virtual Voice good enough for a novel?+
For practical non-fiction, often yes — the job there is clarity. For fiction or memoir, rarely. AI narration is clear but flat: it cannot voice characters, time a joke, or carry emotion, and listeners notice. Virtual Voice titles are also labelled as AI in the store. For voice-led books, a human narrator remains the edition worth publishing.
Where can I sell my audiobook?+
Through ACX your audiobook reaches Audible, Amazon and Apple Books. Going exclusive locks you to those for a higher 50% royalty; non-exclusive lets you also distribute to Spotify, Kobo and libraries via aggregators at 30%. KDP Virtual Voice stays within Amazon, while a recording you produce yourself can be uploaded anywhere you choose.