Most books have never been audiobooks. Recording one meant studio time, a director, an engineer and a narrator working for days, which set a floor under production cost and meant a title had to project a certain sales volume before anyone would attempt it. Frontlist fiction cleared that bar. A twenty-year-old midlist history did not, and neither did most of what any publisher has ever released.
Synthetic narration removes the floor. The consequence is not that audiobooks got cheaper; it is that a category of books which were commercially impossible to record became possible, and the catalogue is expanding into territory that was simply empty.
Two products wearing the same name
What has emerged is a split, and it is cleaner than the argument around it suggested. For a reference text, a technical manual, a procedural non-fiction title or a backlist book somebody wants in the car, a competent synthetic read is sufficient — the listener wants the information, and the narration is a delivery mechanism. Publishers report this material performing perfectly well.
For anything where the reading is part of the work, it is not close. A novel with several characters, a memoir in the author's voice, comic timing, a narrator who understands a sentence before saying it — these are performances, and the gap is audible immediately to anyone listening for it. The market has sorted accordingly, and the sorting happened faster than either the enthusiasts or the alarmists expected.
The economics of the backlist are the genuinely new thing. A publisher sitting on decades of titles now has a route to monetising them at a unit cost that makes the long tail viable, which is the same discovery arriving in a different medium as the archive turning out to be the asset. Rights are the constraint rather than technology: a great many older contracts do not clearly convey audio rights at all, and clearing them costs more than producing the recording.
For narrators, the effect is a compression of the middle. The top of the profession is fine and arguably better paid, because performance is now the explicit thing being bought. The bottom — straightforward non-fiction, corporate work, anything where a clean read was the whole job — has largely gone. What has vanished is the reliable mid-tier work that made narration a career rather than a sideline, which is a familiar shape: the entry-level work that used to train everyone, removed from another profession.
The disclosure question is unresolved and will not stay that way. Some platforms label synthetic narration, some do not, and listeners who care about the distinction currently have no consistent way to act on it. Publishers arguing against labelling are making a short-term case; the ones treating a human performance as a feature worth advertising are describing the market that actually exists, in the same way print returned as a premium object once the cheap version became universal.
Topics mediapublishingai
