Selling text-heavy games overseas: visual novels and JRPGs

If you make visual novels, adventure games, or story-driven RPGs, you have probably had the same conversation twice. Someone asks whether you will release in English, or Chinese, or both. You look at your script — a few hundred thousand characters, maybe more once the branches are counted — and the answer stalls, because everything about the cost of translating scales with exactly the number your genre is largest in.

The awkward part is that the same number cuts the other way too. Story games travel unusually well. Their appeal is characters and writing, both of which survive a border intact, and the people who play them are readers by definition — the audience most willing to spend twenty hours with a text box is also the audience that goes looking for imported titles in the first place.

This article is about the business decision rather than the implementation: how to get a real word count, how to stage the work so you are never betting everything at once, which markets are actually worth adding for this genre, and what goes wrong when text-heavy games are localized badly.

The paradox of text-heavy genres

For most genres, localization is a line item. For yours, it can approach a second development budget. An action game might carry a few thousand words of UI, item names, and tutorial prompts, and its quote reads like a manageable expense. A mid-length visual novel can carry a hundred times that, and the same per-word rate produces a number that stops the conversation.

At the same time, this is the genre where the language decision changes the outcome most. A puzzle game that is not translated still sells to people who cannot read it, because the mechanics carry it. A narrative game that is not translated does not sell to that audience at all — there is nothing left to enjoy when the words are inaccessible. The product is the text.

That combination is why the framing 'should we localize' rarely helps here. The costs are too large to swallow casually and the upside is too structural to dismiss, so the useful questions are different ones: how much, in what order, and paid for out of what. Everything below is aimed at those.

One more thing works in your favor, and it is easy to overlook. Story games have long commercial lives. They keep selling through seasonal sales, bundles, back-catalog discovery when you release your next title, and recommendations inside reading communities that stay active for years. A translation is a one-time cost sitting in front of a very long tail, which makes the payback arithmetic different from a game that does most of its business in launch month.

Get your real number before deciding anything

Nobody can plan from 'a lot of text.' Before you talk to anyone about money, extract the script and count it, because the number is usually not what the team assumes — sometimes lower, often higher once branches are included.

Count everything that ships, not just the dialogue. Narrative lines and narration, every choice and every branch, system and UI text, item and glossary entries, achievement names and descriptions, tutorial text, credits, and store copy. Then look for the text that will not appear in any export: words baked into background images, signs, title logos, and UI art. Those are invisible to a word count and expensive per word, and finding them late is how a budget slips after it was approved.

Then clean the number. Deduplicate repeated system messages and lines that appear in more than one route, because some engines store a physical copy per branch. Exclude cut content still sitting in the files, debug strings, and placeholder text. What remains is your shippable volume.

Finally, resist the conversion ratios that circulate online for turning Japanese characters into English words. They are averages of other people's games, and dialogue-heavy Japanese with short lines behaves differently from dense narration. Get a representative sample translated properly — a few thousand characters covering dialogue, narration, and UI — and measure your own ratio from it. That sample costs little, gives you a number you can defend, and doubles as an audition for the translator you are considering.

Stage it: store page, then a prologue, then the script

The reason staging matters more in this genre is that the full commitment is so large. Each stage below costs a fraction of the next one and produces evidence about whether the next one is justified.

The first stage is the store page and an honest supported-languages declaration. A few hundred words, translated properly by a native speaker, plus a localized screenshot set. This tells you whether anyone in that language is looking at all, and it works even while the game itself is untranslated — as long as the page says plainly what is and is not available.

The second stage is where your genre has a real structural advantage over others: a prologue, a first chapter, or a demo. Narrative games come with natural cut points, and a translated prologue is a complete experience rather than a fragment. It gets streamed, it gets reviewed, it produces wishlists you can count by region, and it costs a small percentage of the full script. No other genre gets to run its market test this cheaply on this much real content.

The third stage is the full script, and here the important decision is not who translates but who holds it together: one lead maintaining a glossary of names, places, and recurring terms, plus character voice notes, from the first line onward. If you split the work across several translators without that role filled, you will pay for consistency twice — once in the translation and again in the review pass that finds the drift.

One warning about staging in public. Shipping half a story translated is worse in this genre than shipping none, because players stop at the exact point the translation does and review from there. If you release chapter by chapter, announce the schedule up front, put it on the store page, and hold to it.

The market you would add is not always the biggest one

The overall language distribution of a storefront's users is a poor guide for narrative games, because genre audiences cluster differently from the population as a whole. The languages where your kind of game has the deepest readership may sit well below the top of a general ranking, and the reverse happens too.

Three sources will tell you more than any general statistic. First, the review-language distribution on comparable titles — games near yours in genre, length, and price that already shipped multiple languages. Second, where your own regional wishlists and traffic come from today, despite the absence of a translation; interest that survives a language barrier is unusually strong evidence. Third, which languages the fan communities around your genre actually discuss in, which is visible for free in the places those discussions happen.

English deserves separate treatment, because in this genre it functions as a hub rather than just another market. Press coverage, streamer discovery, and community discussion largely happen in English, and — importantly — further translations are frequently produced from the English version rather than from your original. That makes English quality compound: an error or a flattened character voice in the English script propagates into every language translated from it.

If you do use English as a relay language, decide it deliberately rather than by drift. Tell the downstream translators that they are working from a relay, keep the original script accessible to them for disambiguation, and expect to answer questions from two directions at once. And where a language has meaningful regional splits — Simplified versus Traditional Chinese, Latin American versus European Spanish — treat that as a real choice with its own audience rather than a formatting detail.

The money side: a one-time cost against a long tail

Text-heavy games rarely earn back a full translation in launch month, and judging the investment on launch month is the most common way teams talk themselves out of a decision that would have worked. The relevant horizon is years, because that is how long these games sell.

There is also a compounding effect across titles. The first translation you commission buys more than a translated game: a glossary, a translation memory, a translator who now knows your world and your characters, and an audience in that language who will be there when your next game appears. For a studio that intends to keep making games in this genre, the first localization is partly an investment in a pipeline, and the second one is meaningfully cheaper than the first.

Pricing deserves a deliberate pass rather than an automatic conversion. A long reading game is bought by people in markets with very different purchasing power, and regional prices set by mechanical currency conversion can put your game out of reach in exactly the markets where your genre has the most enthusiastic readers. Look at what comparable localized titles charge in those currencies.

Finally, budget the recurring part. Every patch, added route, and typo fix in the original script has to travel through every language you have shipped. Revising the source text after translation has begun is the single most reliable way to multiply a localization budget, which is why text freeze discipline matters more here than in any other genre.

What goes wrong in this genre specifically

The failure modes below are not general localization problems. They are the ones that show up disproportionately when the script is the product.

• Translation starts before the script is final, so every later revision multiplies across languages and quietly re-breaks consistency
• A long script is split across translators with no lead, no glossary, and no character voice notes, producing drift that readers notice and reviewers name
• Text boxes, fonts, and line limits are sized around the original language, and expansion breaks layout somewhere deep in a route nobody replays before release
• Name order, honorifics, and key terminology are decided ad hoc per chapter instead of once, at the start, where they belong
• The bulk is machine-translated and only the important parts are checked by hand — in a game whose product is text, every part is an important part
• A partial translation ships without saying so, and players discover the boundary at the moment their experience stops working
• The decision waits for proof that never arrives, because the game cannot sell to an audience that cannot read it