Literary translation re-creates a novel's meaning, voice, and effect in another language, not just its words. Unlike literal or machine translation, it preserves tone, idiom, dialogue register, and a consistent canon of names across a series, so the translated book reads as if it were written in the target language.
You finished the series. Readers in your own language tell you the world feels real and the characters feel like people they know. Then a foreign-rights opportunity appears, or you decide to go wide into a market you cannot read, and you reach for the cheapest thing labeled translation. That is where it goes wrong. The word translation hides two completely different jobs, and authors who confuse them ship an edition that reads like a manual: technically correct, emotionally dead, with the invented city you named so carefully spelled three different ways across three chapters.
The fear underneath this is real and specific. You are about to hand your book to a process you cannot proofread, in a language you do not speak, and trust that your voice survives the crossing. A reader in that market who buys all six volumes can see the seams you never will. So the first thing to get right is not a tool or a price. It is the distinction between literal translation, machine translation, and literary translation, because only one of the three is trying to do what your book actually needs.
Literary translation is the re-creation of a work of fiction in another language so that it carries the same meaning, voice, and emotional effect as the original, rather than the same dictionary words. A literal translation answers the question what do these words say. A literary translation answers a harder one: what does this passage do to a reader, and how do I make it do that again in a language with different rhythms, different idioms, and a different culture sitting behind every line.
That difference is everything in fiction. A threat has to still chill. A joke has to still land. A formal mode of address that signals a character's deference has to keep signaling it. And the name you invented for your protagonist, the order of knights she belongs to, the curse her people use, all have to read the same way every single time they appear, across a whole series, even when years and different translators separate one book from the next. A translation that gets the words right and loses all of that is not a translation of your novel. It is a paraphrase of its plot.
The three approaches share a name and almost nothing else. The table below sets them side by side on what actually matters to a novelist taking a series to a new market.
| What matters | Literal / word-for-word | Generic machine translation | Literary translation |
|---|---|---|---|
| Unit of work | The word and the grammar rule | The sentence, in isolation | The effect of the whole passage |
| Voice and tone | Flattened to plain meaning | Inconsistent, drifts by sentence | Preserved as a deliberate choice |
| Idiom and metaphor | Rendered word-for-word, often nonsense | Sometimes literal, sometimes lost | Re-cast so the figure still works |
| Invented names and terms | Treated like any common noun | Re-translated from scratch each time | Held to one fixed form everywhere |
| Dialogue register | Ignored | Guessed, no memory between lines | Attributed and kept consistent per speaker |
| Across a series | No cross-book memory | No cross-book memory | One canon shared across every volume |
Generic machine translation, the engines built into Google Translate or DeepL, sits in the middle of that table and is built for documents, not novels. It translates each sentence on its own, with no memory of the harbor town it rendered two pages ago. That is fine for a contract. It is fatal for a book whose whole texture is built from words it has never seen and never will: the names you made up. Literal translation sits at the far end, faithful to the words and deaf to the music. Only literary translation is trying to carry the book itself, not its sentences, into the new language.
Most writing means roughly what it says. Fiction hides meaning in things a dictionary cannot hold, and those are exactly the things that break in translation. Four categories cause most of the damage.
The names you coined have no dictionary entry, so a translator has to decide whether to transliterate, adapt, or keep them in the original. Whatever the choice, it has to be the same in book one and book four. A capital transliterated in one volume and left alone in the next is the first thing a reader who bought the whole series notices.
An idiom translated literally is gibberish; a voice translated flatly is gone. Many languages also encode social distance in the words for you and in the titles characters use for each other. When a character who always addresses the queen formally suddenly uses the familiar form, that is a character change the author never wrote.
A reference that lands instantly for one culture can mean nothing, or the wrong thing, in another. Literary translation makes a judgment call: keep it, adapt it, or replace it with something that produces the same reaction. A document translator never has to make that call. A novelist's translator makes it on nearly every page.
Books come out years apart, the translator who did book one may not do book four, and even the same translator does not remember exactly how they rendered a coined term three volumes ago. The same name, place, or term ends up rendered slightly differently each time, in a language the author cannot read to check. The series fractures precisely where no one is watching.
For a long time the honest answer to can software do literary translation was no, not really. That has changed for one reason: the difference between a tool that translates sentences and a tool that knows your book. StoryHelm's Translation Studio is the second kind. Before it renders a single line, it reads your Canon, the structured record of characters, places, factions, and items that StoryHelm has already cataloged from your manuscript, and it holds those names to one consistent form everywhere they appear.
None of that pretends to be a human's final judgment about whether a joke still lands or a threat still chills. What it does is take over the mechanical consistency and entity awareness that machines are genuinely good at, and that ordinary translation gets wrong, so the human judgment is spent on craft instead of on bookkeeping. For comparison, a human literary translator typically costs ten to fifteen thousand dollars per book per language, which puts a six-book series in five markets out of reach for most indie authors. Translation Studio starts at one hundred ninety-nine dollars per language, one time, and you own every word of the result.
You wrote the finished book. Translation Studio carries it into other languages, it does not write your original creative prose. Because it is anchored to the Canon you already built, it keeps every character name, place, and invented term steady across the whole series, scores each chapter for publishability before you download, and back-translates to verify the meaning held. The author wrote the story; Translation Studio renders it faithfully, and you own the translated output completely, with no royalties and no co-authorship claims.
Literary translation re-creates a work of fiction in another language so it carries the same meaning, voice, and emotional effect as the original, not just the same words. It preserves tone, idiom, dialogue register, and cultural nuance, and holds invented names and terms steady, so the translated novel reads as if it were written in the target language.
Literal translation maps words and grammar across languages sentence by sentence, which flattens idiom and voice. Literary translation reproduces the effect: the rhythm of the prose, the register of dialogue, the weight of a metaphor, and a consistent canon of invented names. Literal aims for accuracy of words; literary aims for fidelity of experience.
AI can do far more than literal translation when it is entity-aware. StoryHelm's Translation Studio reads your manuscript canon, keeps character names, places, and invented terms steady across the whole series, attributes dialogue, supports regional dialects, scores every chapter for publishability, and back-translates to verify meaning. A skilled human still makes the final literary judgment calls.
Fiction carries meaning in things a dictionary cannot hold: invented names, idiom, voice, dialogue register, cultural nuance, and facts that must stay consistent across hundreds of pages and many books. A document translator renders each sentence in isolation. Fiction needs every recurring name and term held to one form across the entire series.
StoryHelm renders your story with its voice, idiom, and canon intact across 51 languages and dialects, and scores every chapter for publishability before you download.
Request a founding seat