Why self-publishers should value human qualities
These days, artificial intelligence tools like DeepL can deliver translations in seconds—and they’re getting better all the time. Sometimes I wonder: will the machine soon be able to translate better than I can? The progress is impressive, especially with clear, factual texts. But when it comes to literature—books full of emotion, imagery, and subtle undertones—this is where the wheat is separated from the chaff. Literary translation is more than just transferring words. It’s about capturing emotions, reading between the lines, and creating that movie in the reader’s mind.
AI in Non-Fiction Translation
I recently read that some publishers are now using AI to translate non-fiction books. These texts are then revised by human translators or editors. This involves eliminating errors and refining the style. Although non-fiction is primarily about conveying facts, these texts should still be entertaining and written in a relaxed style. To achieve this, translators must detach themselves from the source text and adopt a freer approach. Artificial intelligence adheres slavishly to the original and, at worst, produces a dry, stiff text. What should inform and help people (non-fiction books on health topics) will just gather dust. However, AI inevitably produces such texts because no machine—and that is precisely what artificial intelligence is—has a sense of humour or can read between the lines. Even if the author lightens a serious topic with a funny anecdote, there is a high probability that the tone will be lost in translation. After all, robots don’t tell jokes.
What About Fiction?
Many literary translators, even those with decades of experience, are seeing fewer assignments. It is quite possible that previous clients are using AI here, too. Translators don’t work for free, but AI does the job for next to nothing.
Even self-published authors are using programs designed for writers to translate their books into all sorts of languages—supposedly with convincing results. I see it differently, but maybe I’m just too demanding. In my opinion, the sample chapters of these AI-translated books were interchangeable and just “okay.”
A good book hooks the reader within the first fifty pages. The opening sentences often decide whether a book is bought or read at all. If I’m not immersed in the story after four chapters, that’s it—it goes on the shelf and gathers dust. What a waste of paper.
What Makes a Book Stand Out?
Of course, that’s subjective, but it usually comes down to this: an engaging plot, characters you can relate to, and a lively, flowing style that moves you, stirs emotions, and creates vivid images. An author who follows the “show, don’t tell” principle—showing feelings and actions instead of just stating them—is on the right track. But not everyone does that. Sometimes you’ll see, “He was angry,” instead of, “He clenched his fists and gritted his teeth.” Or, “She was so annoyed,” instead of, “She rolled her eyes and pulled a face.” An AI will translate exactly what’s on the page—nothing more, nothing less. A skilled literary translator, who knows how to play with imagery and emotion and read between the lines, will get much more out of such a text. Even the best plot falls flat without imagery and interpretation.
See for Yourself
Let’s look at a passage from O. Henry’s classic short story, The Gift of the Magi.
Original:
Della finished her cry and attended to her cheeks with the powder rag.
She stood by the window and looked out dully at a gray cat walking a gray fence in a gray backyard. Tomorrow would be Christmas Day, and she had only $1.87 with which to buy Jim a present. She had been saving every penny she could for months, with this result. Twenty dollars a week doesn’t go far. Expenses had been greater than she had calculated.
They always are. Only $1.87 to buy a present for Jim. Her Jim. Many a happy hour she had spent planning for something nice for him. Something fine and rare and sterling—something just a little bit near to being worthy of the honor of being owned by Jim.
There was a pier glass between the windows of the room. Perhaps you have seen a pier glass in an $8 flat. A very thin and very agile person may, by observing his reflection in a rapid sequence of longitudinal strips, obtain a fairly accurate conception of his looks. Della, being slender, had mastered the art.
Suddenly she whirled from the window and stood before the glass. Her eyes were shining brilliantly, but her face had lost its color within twenty seconds. Rapidly she pulled down her hair and let it fall to its full length.
DeepL’s German Translation:
(Back-translated for illustration)Della stopped crying and wiped her cheeks with the powder rag.
She stood at the window and looked sadly at a gray cat walking along a gray fence in a gray backyard. Tomorrow was Christmas, and she had only $1.87 to buy Jim a present. She had saved every penny for months, and this was the result. Twenty dollars a week didn’t go far. The expenses had been higher than she thought.
They always are. Only $1.87 to buy a present for Jim. Her Jim. She had spent many happy hours planning something nice for him. Something fine and rare and valuable—something that was a little worthy of belonging to him.
There was a pier glass between the windows of the room. Perhaps you have seen a pier glass in an $8 apartment. A very slim and very agile person can, by looking at their reflection in a quick sequence of longitudinal strips, get a fairly accurate idea of their appearance. Della was slim and had mastered this art.
Suddenly she spun away from the window and stood in front of the mirror. Her eyes shone brightly, but her face had lost its color within twenty seconds. She quickly pulled down her hair and let it fall to its full length.
My Literary Translation:
After Della’s tears had dried, she dabbed her cheeks with the powder puff.
Listlessly, she gazed out the window, watching a gray cat balancing along a gray fence in a gray backyard. Tomorrow was Christmas, and she had only one dollar and eighty-seven cents to buy Jim a present. Even though she’d pinched every penny for months, there was hardly anything left. Twenty dollars a week just didn’t stretch far enough. And her expenses had, as always, been higher than expected.
That was always the way. A measly one dollar eighty-seven for her Jim! Yet she had spent hours, blissfully planning something tasteful, unique, and precious—something that would show Jim what an honor it was to belong to him.
Do you know those narrow mirrors you find in a typical eight-dollar apartment? There was one between the windows. If you’re slender and quick, you can catch a glimpse of your reflection and get a sense of how you look. Della, being slim, had mastered this art perfectly.
She turned abruptly from the window and positioned herself in front of the mirror. Staring back were bright eyes in a face that had lost all its color in a split second. Hurriedly, she let down her long hair, hiding her pale face behind it.
Why the Difference Matters:
To highlight Della’s drama and dilemma, I chose to be a bit freer in places. The reader should be able to feel with Della. DeepL translates the text correctly and without errors, but it’s too literal. Is it vivid? Does it create images and emotions? Not really. That’s where the human touch makes all the difference.
How Does AI Handle Tricky Cases?
Idioms and typos are often recognized by AI these days—there’s been real progress here. But what about less obvious mistakes? Words that sound similar, where a single letter changes the meaning?
Let’s take a look:
- She excepted his invitation to dinner.
- Should be “accepted“
- DeepL: “Sie lehnte seine Einladung zum Abendessen ab.”
- Close, but not quite. She accepted the invitation, not rejected it.
- They complimented each other on their new suits.
- Should be “complemented” = to go well together
- DeepL: “Sie haben sich gegenseitig zu ihren neuen Anzügen beglückwünscht.”
- Why would you congratulate someone on a suit? That’s odd.
- The dessert was beautiful at sunset.
- “Dessert” instead of “desert”
- DeepL: “Das Dessert sah bei Sonnenuntergang wunderschön aus.”
- When has a dessert ever influenced the plot of a novel? Unless it’s poisoned, but that’s not the case here.
A human translator, paying attention to context, would have caught these small errors. Artificial intelligence doesn’t “think” or read context. That’s why it’s still a machine, not a person.
Metaphors, Similes, and Wordplay
Even though translation tools now often recognize idioms, they still struggle with tricky metaphors, similes, and wordplay. Literary translators wrestle with these all the time, often having to invent new expressions in the target language. That takes real brainpower—something no machine can supply.
Metaphors and comparisons that work in English can sound odd or even ridiculous in German (and vice versa). Sometimes, you have to find a new image that fits the context and conveys the same feeling. You might spend weeks polishing a single phrase, coming back to it again and again until it finally works. Sometimes, you have to let a metaphor go. But you can make up for it elsewhere, creating a stronger image or emotion than in the original. It’s all about weighing your options—a decision human translators make with every book. DeepL, on the other hand, just churns out odd sentences.
Examples? Here you go:
- Her thoughts fluttered around her head like drunken butterflies.
- DeepL: “Ihre Gedanken flatterten wie betrunkene Schmetterlinge um ihren Kopf herum.”
- How about: “Her thoughts spun in dizzying circles, making her head swim.”
- He moved through the crowd like a rumor—unseen but felt.
- DeepL: “Er bewegte sich durch die Menge wie ein Gerücht – unsichtbar, aber spürbar.”
- How about: “He slipped through the crowd like a shadow—no one saw him, but everyone sensed his presence.”
Again, the AI translates literally or smooths out the image, losing the literary effect. This is where you need creativity, the courage to break away from the source text, and the instinct to find your own solution.
Wordplay and Idioms
Wordplay is another classic stumbling block for AI. For example, take the idiom “kick the bucket” (to die). If an author writes, “The farmer kicked the bucket, and now his cows are thirsty,” a machine might translate it as “The farmer died, and now his cows are thirsty,” missing the joke entirely. Or worse, it might translate it literally, making no sense at all.
Authors like O. Henry, Mark Twain, Oscar Wilde, and Lewis Carroll are famous for their love of puns and idioms—another reason why literary translation will always need a human touch.
Maybe translation AIs will get even better in the future. But as long as literature is more than just information—as long as it aims to touch, surprise, and move us—human translation will remain irreplaceable. That’s not a weakness; it’s the greatest strength of our profession: we are the ones who read between the lines and bring stories to life.
Feel free to try it yourself: have a machine translate a passage from a novel, then read a literary translation. You’ll feel the difference. Literature needs humanity, creativity, and that famous pinch of intuition. That’s what makes books unforgettable—and literary translation an art.
As long as AI doesn’t get goosebumps from a good sentence, I’ll be happy to keep doing the job.