Should You Translate Your Visual Novel, and Where Would You Start?

A comment arrives on your page in a language you cannot read. You paste it into a translator and it turns out to be someone asking, politely, whether there is a version they can play.

That question is more common than most first-time creators expect, and it is genuinely hard to answer, because the honest reply is neither yes nor no. A translated edition of a visual novel is not a switch you flip. It is a small second project with its own schedule, its own failure modes, and a payoff that varies enormously depending on which language you pick and how the story is written. Text-heavy games are the most expensive kind of game to translate and the kind where translation matters most, which is an awkward pair of facts to hold at once.

So here is the working version, aimed at someone with a finished or nearly finished story, no localization budget worth the name, and a real question about whether this is worth an evening or a month.

When is a story actually ready to translate?

Later than people think. The single most expensive mistake in this whole area is translating text that is still moving.

Every line you change after handoff is a line someone has to redo, and the cost is not proportional to the edit. Rewriting one exchange in the middle of a route means the translator has to reread the surrounding scene to keep the voice consistent, and if there are three languages in flight it happens three times. Lock the script first. Not "mostly done" locked, but the version you would ship if nothing else happened.

There is a real exception. If you know from the start that a translated edition is coming, a few habits during writing make the later work dramatically cheaper, and they cost you nothing now. Keep dialogue out of image files. Avoid puns that carry plot weight. Do not build a scene whose emotional turn depends on an English idiom nobody else has. Write character names so they survive being written in another script.

Which language should come first?

Pick one. Solo creators who announce four languages usually ship zero, and the version that ships is the one where a single language got real attention.

Simplified Chinese is the obvious first candidate on raw numbers. It has traded the top spot with English on Steam's own language survey for a couple of years now, and for visual novels specifically the Chinese-language reading audience is large and active. Japanese is the historical home of the form, with an audience that knows the conventions better than you do, which cuts both ways: high interest, high standards, and a market where English-language releases genuinely do not travel far on their own.

But raw market size is the wrong first filter for a small story. Three things matter more.

Where your readers already are. Look at where your page traffic comes from before you guess. A tiny game with an unexplained cluster of readers in Brazil or Poland has been handed an answer, and Brazilian Portuguese and Polish are both cheaper to commission than Japanese.

Whether you can support the language after release. A translated edition brings comments, bug reports, and questions in that language. If you have nobody who can read them, you have created a channel you cannot answer, and silence reads worse than absence.

Whether the story survives the trip. Heavy wordplay, dense cultural specificity, and jokes built on the shape of English sentences all translate badly and expensively. A dialogue-driven character piece usually travels well. A comedy about a specific regional pun does not.

If you have no signal at all and want a default, Simplified Chinese for reach and Japanese for the form itself are the two that come up most often in indie localization guidance, with Spanish and Brazilian Portuguese as the low-cost, wide-coverage options.

One thing worth doing before any of it: translate the store or project page. It is a few hundred words instead of forty thousand, it is what people read before deciding, and it is the cheapest test of whether an audience is actually there. Steam treats store page languages separately from game languages, and marking a page as available in a language costs nothing but the translation itself.

What breaks when the words change length?

The words are only half the job. The other half is that your interface was built around the shape of the original text, and the new text is not that shape.

Expect European languages to run 15 to 30 percent longer than English, with German regularly past 35. Your dialogue box was sized to the longest English line you happened to write, and now several lines are half again as long. They either overflow, get clipped, or shrink to a font size nobody wants to read.

Chinese, Japanese, and Korean go the other way horizontally and then surprise you vertically. The character count drops, but each glyph is denser and needs more line spacing to stay legible, so a fixed-height box can break in CJK even though the line got shorter. Line breaking is a separate problem: these languages do not put spaces between words, so a wrapping rule that splits on spaces will either never wrap or wrap in the wrong places.

Then there is the font. A typeface designed for Latin script has no Chinese, Japanese, or Korean glyphs in it, and what the player sees is a row of empty boxes. You need a font that actually contains the characters, licensed for embedding, and the free options here are good: Noto and the Source Han family cover CJK properly and are open licensed.

Two more that catch people. Text baked into images, meaning title art, UI buttons drawn as pictures, and any sign or note that appears as part of a background, has to be redrawn or it stays in the original language forever. And variables inserted into sentences break grammar in ways English never warned you about. A line built as "You gave the flower to {name}" assumes English word order; languages that put the verb last, or attach particles to the noun, need the sentence restructured rather than the slot refilled. Where you can, hand the translator whole sentences instead of fragments to assemble.

If you are exporting to Ren'Py, it has a translation framework built in that pulls dialogue out into per-language files, which is one of the better reasons to consider that path for a multi-language release.

Who should do the translation?

The honest hierarchy, cheapest first.

Machine translation alone, for narrative, is not good enough, and 2026 has not changed this as much as the marketing suggests. It handles plain informational text acceptably. It flattens character voice, misses register, resolves ambiguity the wrong way, and translates jokes literally. Readers in the target language spot it in the first few lines, and for a story whose only asset is the writing, that is the whole product failing at the door.

Machine translation with a human editor, usually called post-editing, is the compromise most small teams land on. A translator fixes the output rather than starting blank, and the usual claim is roughly half the cost of a full human pass. It works better on functional text than on emotional text, so it is reasonable for menus, tutorials, and item descriptions, and shakier for the confession scene.

A human translator who plays the game is what you actually want for dialogue, and for a short visual novel it may be less unreachable than you assume. Rates vary widely by language pair, and a 20,000-word story is a smaller job than most people picture. Get quotes before you decide it is impossible.

Then there is the option nobody advertises. Fans offer. Someone who loved your game will sometimes volunteer to translate it, and this can work beautifully or go badly, so treat it like the collaboration it is: agree on scope up front, agree on credit, agree in writing that you can use the result commercially, and be prepared for it to stall halfway. Do not accept a full-script translation from someone you cannot verify and then ship it unread.

Whoever does it, give them context. A character sheet with each speaker's age, relationship to the others, and how formally they talk. A glossary of names, places, and any invented terms with the spelling you want. Notes on jokes that need adapting rather than translating. Access to the actual game so they can see who is on screen. Translators working from a bare spreadsheet of lines produce bare spreadsheet results, and it is not their fault.

How do you check work you cannot read?

You do not proofread it. You get someone to play it.

One native reader playing the translated build start to finish will find more than any amount of staring at the text file. They catch the line that is technically correct and tonally wrong, the honorific that makes two characters sound like strangers, the choice button that got clipped, the name spelled two different ways in two routes. Pay them if you can. Trade a favor if you cannot. Do not skip it.

Alongside that, a short pass of your own catches the mechanical failures without knowing a word of the language. Play the whole thing and look at shapes, not meaning: text overflowing the box, empty rectangles where glyphs should be, lines wrapping mid-character, buttons whose labels no longer fit, dates and numbers in the wrong format, and any place where the original language is still sitting there because it was inside an image.

Keep a list of what the reader flags and fix it before release rather than after. A translated edition that ships broken teaches that audience that your work is not for them, and they will not check back.

How does Novelez change this?

Novelez does not translate anything for you, and any tool that claims to do this well without a human in the loop is selling you the version of the problem described above.

What changes is the mechanics around the text. Your script lives inside the project as structured scene text rather than scattered through code, so producing the file a translator works from means reading it out of one place, and putting it back means replacing text in the same place. The art, backgrounds, music, and character sprites are already in the project and get reused as they are. A language edition is a copy of the project with the words swapped, not a rebuild.

Because a project publishes to a link rather than a downloadable archive, each language edition is its own link you can hand to that audience, and a correction after release is a fix and a republish. Nobody is sitting on a stale build. That matters more for translation than for anything else, since the first weeks after a translated release are exactly when the fixes arrive.

The free plan holds five projects, which is room for the original and a few language editions without any of this becoming a budget question. Start with one.

What should you check before shipping a translated edition?

• The original script is locked, and no scene is still being rewritten.
• One language, chosen for a reason you can say out loud.
• The store or project page is translated too, not just the game.
• Every text box has been checked at the longest line in the new language.
• The font contains the target script and is licensed for embedding.
• Text inside images has been redrawn, including title art.
• Sentences built from variables read correctly in the new grammar.
• One native reader has played the whole thing and their notes are fixed, not filed.
• Names and invented terms are spelled the same way everywhere.
• You have some way to read the comments that come in after.

A translation is not a bonus feature you bolt on at the end. It is a second edition of the same story, and the ones that land are the ones treated that way. One language, done properly, will reach more people than four done at speed.