Audiobooks narrated by Amazon's virtual voice pay authors a 40 percent royalty, while human-narrated titles on Audible's new exclusive terms pay 50 percent. The machine option earns the lower rate.
That inversion is recent, and most cost comparisons have not caught up with it. For years the pitch was simple, since skipping a narrator removed the biggest line in the budget and nothing else changed. The royalty half of the equation has now moved against the synthetic option.
The production savings are still enormous. A ten-hour audiobook runs roughly 400,000 characters, which costs somewhere between $20 and $40 to generate at current per-character rates. A human narrator charging $150 to $400 per finished hour costs $1,500 to $4,000 for the same book, so the underlying cost of machine-made speech is two orders of magnitude cheaper before anyone argues about quality.
Run the arithmetic on the top of that band. A $14.99 title at 40 percent returns about $6 a sale before Amazon adjusts the selling price, which it reserves the right to do independently of what you set.
The 40 percent figure is quoted for a la carte sales as well, and Audible's own royalty structure is now built around membership value rather than single purchases.
The catalog rules are narrow too. Virtual voice remains a beta limited to the US marketplace, titles are clearly labeled as machine-narrated, and the voice choice covers around 80 options across American, British and Australian English, Latin American and Castilian Spanish, French and Italian.
You can vary the voice by chapter, which helps with anthologies and multi-narrator fiction. Books enrolled in KDP Select also enter Audible's Plus catalog, paid from a separate fund on top of Kindle Unlimited earnings.
Titles that are not enrolled have to stop distributing. That applies to backlist titles you published years ago and have not thought about since, which is the category most likely to be missed.
The calculation changed shape as well, not just the percentage. Payment is based on Member Value, meaning the price of a listener's monthly membership minus taxes and fees, multiplied by your contractual rate rather than by a single sale price.
When a member spends extra credits, that value gets added in and divided across titles in proportion to their à la carte price. Listing price still matters, but it now sets your share of a pool rather than your cut of a transaction.
Enrolled titles also gain suggested pricing and the option to join the All You Can Listen program. Both are earnings levers that did not exist under the old terms.
Translation is different because the cost was previously impossible rather than merely high. Recording the same title again in four languages meant four narrators, four sessions, and four sets of direction, which almost no backlist could support.
Audible now offers publishers translation from English into Spanish, French, Italian and German, running either as text-to-text on the manuscript or as speech-to-speech that preserves the characteristics of the original narrator. Publishers can add professional linguist review and edit the translation before it ships.
The second route is the one that changes producing an audiobook without a narrator into something else entirely. A title already recorded once by a human can reach four more markets while keeping the performance that made it work, without booking anybody again.
Publishers get more than 100 machine voices for straight narration too, spanning English, Spanish, French and Italian, through either a managed service where Audible runs the whole production or a self-service path using the same technology.
The binding constraint on all of it is policy rather than price. Where you can publish a synthetic narration, how it must be labeled, which languages are supported, and which marketplace you are limited to all sit outside your control, and they change more often than the generation cost does.
That inversion is recent, and most cost comparisons have not caught up with it. For years the pitch was simple, since skipping a narrator removed the biggest line in the budget and nothing else changed. The royalty half of the equation has now moved against the synthetic option.
The production savings are still enormous. A ten-hour audiobook runs roughly 400,000 characters, which costs somewhere between $20 and $40 to generate at current per-character rates. A human narrator charging $150 to $400 per finished hour costs $1,500 to $4,000 for the same book, so the underlying cost of machine-made speech is two orders of magnitude cheaper before anyone argues about quality.
The price ceiling caps what a synthetic title can earn
Amazon sets hard limits on what these titles can sell for. A virtual voice audiobook lists between $3.99 and $14.99, and no amount of polish moves it above that ceiling.Run the arithmetic on the top of that band. A $14.99 title at 40 percent returns about $6 a sale before Amazon adjusts the selling price, which it reserves the right to do independently of what you set.
The 40 percent figure is quoted for a la carte sales as well, and Audible's own royalty structure is now built around membership value rather than single purchases.
The catalog rules are narrow too. Virtual voice remains a beta limited to the US marketplace, titles are clearly labeled as machine-narrated, and the voice choice covers around 80 options across American, British and Australian English, Latin American and Castilian Spanish, French and Italian.
You can vary the voice by chapter, which helps with anthologies and multi-narrator fiction. Books enrolled in KDP Select also enter Audible's Plus catalog, paid from a separate fund on top of Kindle Unlimited earnings.
Audible's new royalty model has a hard deadline
The rate change is not optional, and it is running on a clock. Enrollment in the new model opened to all ACX rights holders on 26 May 2026, and the legacy model is discontinued by the end of the year.Titles that are not enrolled have to stop distributing. That applies to backlist titles you published years ago and have not thought about since, which is the category most likely to be missed.
The calculation changed shape as well, not just the percentage. Payment is based on Member Value, meaning the price of a listener's monthly membership minus taxes and fees, multiplied by your contractual rate rather than by a single sale price.
When a member spends extra credits, that value gets added in and divided across titles in proportion to their à la carte price. Listing price still matters, but it now sets your share of a pool rather than your cut of a transaction.
Enrolled titles also gain suggested pricing and the option to join the All You Can Listen program. Both are earnings levers that did not exist under the old terms.
Translation is where the numbers actually move
Narration was never the hardest economics problem. A book that sells enough to justify $3,000 of studio time was always going to get recorded, and the titles that did not sell enough simply stayed silent.Translation is different because the cost was previously impossible rather than merely high. Recording the same title again in four languages meant four narrators, four sessions, and four sets of direction, which almost no backlist could support.
Audible now offers publishers translation from English into Spanish, French, Italian and German, running either as text-to-text on the manuscript or as speech-to-speech that preserves the characteristics of the original narrator. Publishers can add professional linguist review and edit the translation before it ships.
The second route is the one that changes producing an audiobook without a narrator into something else entirely. A title already recorded once by a human can reach four more markets while keeping the performance that made it work, without booking anybody again.
Publishers get more than 100 machine voices for straight narration too, spanning English, Spanish, French and Italian, through either a managed service where Audible runs the whole production or a self-service path using the same technology.
The binding constraint on all of it is policy rather than price. Where you can publish a synthetic narration, how it must be labeled, which languages are supported, and which marketplace you are limited to all sit outside your control, and they change more often than the generation cost does.
