Audiobooks allow readers to move through the world, accompanied by a story. But how would the quality of the story change if that voice originated not from a fellow human, but from a machine? On May 13, Audible sowed the seeds of a widespread version of this reality by announcing that the company will soon be using AI to narrate and translate select Audible books into languages including English, Spanish, French, and Italian. The news arrives as the book industry scrambles to keep up with fast-moving AI technologies, which are already being used by some companies—such as the Simon & Schuster subsidiary Veen Bosch & Keuning—to translate commercial fiction.
Unlike some virtual “text-to-speech” services, AI-generated voices are an algorithm-driven aggregate of real-life voices studied for cadence, tone, and emotion, among other distinctly “human” qualities. Delivered at rates that can be as low as a tenth of what human voice actors charge for book narration, AI narrations have been quietly improving—and becoming nearly indistinguishable from human voices—in a trend that leaves some in the industry unnerved.
“Storytelling is one of the oldest art forms. It’s how we teach what it means to be human, and I think it’s so weird that people would want a computer to tell them a story,” says Henrietta Meire, a professional voice actor who has narrated over three hundred fifty audiobooks to date. For Meire, narration is an “immersive experience” that requires close reading, extensive research, and the ingenuity to generate voices for hundreds of characters, as she did for a fantasy series. By her account a narrator is to an audiobook what a translator is to an original text: a participant in the creative process whose indelible human fingerprint is left on the work itself.
For many authors this fingerprint is especially prominent. “As a writer and a listener of audiobooks, I always prefer audiobooks that are read by the author because they are reading it in the way that they intended it to be heard,” says journalist and nonfiction writer Alex Cuadros, who recently narrated his own book When We Sold God’s Eye (Grand Central Publishing, 2024). In weaving non-English words into English prose, books like When We Sold God’s Eye—which follows the Indigenous Cinta Larga tribe in Brazil’s northwestern state of Rondônia—contain deft transitions between voices and languages that not only benefit from an author’s linguistic repertoire and lived experience, but also risk being seriously flattened by AI.
Lauren Elkin, a French and American writer who most recently narrated the transatlantic and multilingual world of her first novel, Scaffolding (Farrar, Straus and Giroux, 2024), decided to record the audiobook—one that encompasses her familiarity with both French and English—for the ability to create a culturally distinctive universe that echoes her own. “I wish [AI] would be limited to fields where it can make a quantifiable contribution to human endeavor and not replace it altogether,” she wrote in an e-mail. “I will cop to occasionally feeding a hard-to-parse sentence into Google Translate just to see what it thinks, but it’s only helpful a fraction of the time—far more useful is just getting out the old Larousse [dictionary].”
Ilan Stavans, the publisher of Restless Books, offers yet another perspective. He acknowledges the potentially harmful effects of AI while also understanding that AI-driven narrations can offer a valuable service to some voice actors and publishers.
Having recently worked with a voice actor to pronounce a variety of Yiddish accents for an anthology he edited, How Yiddish Changed America and How America Changed Yiddish (Restless Books, 2020), he suggests that a narrator could use AI as a helpful tool to learn how the same word in one language is pronounced differently in various regions. “We want to be as attentive to the nuances of sound as we are attentive to the nuances of script on that page,” he says.
Despite her reservations, Meire is similarly open to AI’s possibilities. Recognizing that some narrators choose to read under pseudonyms for assignments that are financially necessary but not in the narrators’ desired genres, she wonders if the ability to license and clone one’s voice—the cloning process can cost anywhere from $8,000 to $12,000, to Meire’s knowledge—would allow narrators to prioritize the jobs that they are most excited about. “If companies are to increase their use of AI-generated narrations,” she says, “there need to be protections in place for narrators to be able to own their clones, license them out, and issue fines to people who steal voices.”
Prior to the publication of When We Sold God’s Eye, Cuadros visited the Cinta Larga people with an AI-generated translation of his book so that they could “know for themselves what was in it, rather than hear about it secondhand or through media coverage.” The gesture—one intended to convey respect and inclusion and made possible by the rate at which AI works—allowed the Cinta Larga to consider Cuadros’s book before the rest of the world could.
But Cuadros has no doubt that, should his book be published in Portuguese, he wants a human translator for the job. “There’s so much thought and care that goes into considering the weight of a word, the rhythm of a sentence, and I don’t think that AI will ever be able to replicate that,” he says.
Stavans is quick to agree. For him, the interpersonal is paramount: The most minute conversations about accents, pronunciations, and intercultural meanings he shares with narrators, authors, and producers are “the gist” of why he does what he does. Though an AI-generated narrator might be able to pronounce the words correctly, “the passion, the intention, the pathos will not be there,” he says. “It’s like saying you can do AI theater with robots. They can pronounce the lines in Death of a Salesman, but there will not be a soul in that performance.”
Jimin Kang is a Seoul-born, Hong Kong–raised, and England-based journalist and writer. Her work has appeared in the New York Times, the Nation, the Kenyon Review, and the Los Angeles Review of Books, among other publications.
Thumbnail credit: and machines on Unsplash