Translation can feel like a daunting leap for independent publishers. Traditional literary translation is both expensive and slow, leaving many projects stalled before they start. AI-assisted translation promises to lower those barriers, but questions remain about quality, cultural nuance, and the role of the translator.
To get a clearer picture, IBPA Independent spoke with Yves Vermeulen, chief operating officer and co-founder of Nuanxed, and Dr. James Hadley, Ussher associate professor in literary translation and director of the Trinity Centre for Literary and Cultural Translation at Trinity College Dublin. Both agree that while AI can add efficiency, human oversight remains essential.
Lowering the Threshold for Indies
Independent publishers are increasingly curious about AI-assisted translation, Vermeulen says, because traditional approaches often feel out of reach. High costs and long turnaround times can deter publishers from even attempting translation. “Our approach lowers the risk and threshold of translating books,” he says, “which opens up a great opportunity for them to cross borders and attract a wider audience.”
He points to fiction—thrillers, romance, and feel-good titles—as the genres where AI support has proved most useful, though nonfiction is beginning to benefit as well.
A Changing Landscape
Hadley situates the rise of AI within a long history of literary translation practices that have barely changed in more than a century. “The arrival of technology that has the potential to change anything about the scenario is potentially traumatic,” he says. While some publishers are experimenting—large houses on the margins, and some independents more boldly—he notes that “any failure could have substantial reputational, as well as financial, implications,” which makes most companies cautious.
The pressures are real. Literary translation has always been financially precarious, often relying on government subsidies to cover translators’ fees. “Literary translation has remained essentially a cottage industry,” Hadley says. The fear is that if AI is viewed purely as a way to demand more work in less time, “the precarious nature of this situation will only be exacerbated.”
Two Models, Two Philosophies
A central debate lies in how technology is used. Hadley distinguishes between Literary Machine Translation (LMT), where automation dominates and humans “post-edit” machine output, and Computer-Assisted Literary Translation (CALT), where the translator remains in control and AI is just one tool among many.
“In CALT, the human translator retains the agency at the center of the process,” Hadley says, “and makes use of the technology to produce the translation.” This approach, he argues, respects translators’ artistry and avoids the false efficiency of LMT, which often erodes quality.
Nuanxed’s model falls squarely on the CALT side. A vetted translator works through the manuscript, editing and retranslating machine output as needed in a proprietary environment. The text then goes through editing and proofreading, with special collaboration on series to ensure consistency. “AI is just a set of tools to support translators, editors, and proofreaders,” Vermeulen says. “But it is the experienced and vetted linguists who determine the quality of the translation, not the AI involved.”
Pitfalls and Safeguards
Hadley warns that the biggest pitfall is believing the machine can deliver polished literary translations on its own. “By far the biggest pitfall that I perceive is lazy thinking, where people assume the system is a quick fix they can give a text to and have the system produce their ideal translation.”
He stresses that effective use requires detailed “prompt engineering”—instructions that can run to hundreds or even thousands of words, spelling out the intended audience, stylistic goals, and even how to handle puns or sentence length. Without that, he says, outputs are inconsistent and unreliable.
Vermeulen echoes the need for rigor, noting that quality assurance must mirror traditional publishing. Translators are credited for their work and remain accountable for the outcome, while editors and proofreaders provide additional safeguards to ensure a seamless reader experience.
How Translators Are Responding
Translators themselves are divided. Vermeulen reports that most who try Nuanxed’s system are open-minded and see its benefits. “Others are indeed more skeptical and prefer to work in the traditional way, which is perfectly fine as well,” he says. Out of hundreds who have tested the approach, only a handful have stepped back completely.
Hadley sees a generational divide. Established translators, many of whom built careers without formal training, often resist post-editing roles. Emerging translators, by contrast, are more receptive. “They are less likely to have fixed views and philosophies on what translation is and how it works,” he says, and they tend to be digital natives, comfortable experimenting with new tools.
At Trinity, Hadley ensures students are ready for this reality. “Whether people like the idea of using technology to translate literature or not, we think it would be irresponsible to produce graduates who are highly qualified to work in a world that does not exist anymore,” he says. Instead of training them on a specific system, his program teaches them to analyze texts, define audiences, and design detailed strategies—skills that apply whether they translate manually or with AI support.
Looking Ahead
Both experts see AI as a support, not a replacement. “AI is a great starting point, taking away some of the labor-intensive work so translators can focus on the creative aspect of the translation,” Vermeulen says. “It also means more books can be translated, which in turn provides more work to the translator community, not less.”
Hadley cautions that the field is still in flux. He frames AI-assisted translation as moving through a “hype cycle”: breakthrough, excitement, backlash, and eventual integration. “At the time of writing, there are more questions about the potential benefits and challenges than there are solid answers,” he says. What the industry looks like five years from now, he adds, may be very different from anything we can currently imagine.