Gaming

The terrain

Game localization is three different problems that happen to require the same translator.

First: the UI. Menus, HUD elements, button prompts, system messages. This is software localization: character constraints, termbase compliance, TM leverage on recurring strings. The same discipline as any other software product, except the terminology is game-specific (“save slot”, “difficulty”, “achievements”) and the character constraints come from screen layout rather than interface guidelines.

Second: the narrative. Dialogue, item descriptions, lore, journal entries, NPC bark lines. This is literary translation. Register, voice, humor, cultural reference, subtext. A character whose speech pattern marks them as uneducated in English needs a different linguistic register in French: not bad grammar, but a specifically working-class Parisian register. This requires a translator who knows the game world, not just the language pair.

Third: the community. Players read everything and discuss it publicly. A mistranslated ability name becomes a Reddit thread the week before launch. A character whose Japanese voice sounds incongruent with their written dialogue produces forum posts. The quality bar is set not by the developer’s style guide but by the community’s existing expectations, especially for sequels, where fans remember the previous localization.

What we have learned

Register consistency across a long narrative is the hardest problem in game localization and the one most frequently underestimated in project planning.

An 80,000-word RPG translated by four linguists over six weeks will have four different registers unless someone manages consistency actively. The consistency work is not translation. It is editing. Comparing how each translator renders the same character’s speech across the document, finding the breaks, and resolving them to a single voice. This takes time that is not in most project budgets, and the absence of it is visible in the final product.

Humor and cultural reference are the second hardest problem. A joke that lands in English may not translate, not because translation is hard, but because the cultural reference does not exist in the target market. The options are: transliterate the untranslatable joke (confuses players), explain the joke (kills the joke), or replace it with something culturally equivalent that performs the same narrative function. The third option requires a translator with enough creative latitude and enough trust from the developer to make that call.

Audio localization adds a constraint set that text-only localization does not have: lip sync timing, breathing space, and the actor’s delivery. A translated line that reads perfectly may be 40% longer than the English, making it impossible to sync to the existing animation. This requires a back-and-forth between linguist and audio director that is different from standard localization review.

A typical engagement

We start with a glossary and style guide: the game’s core terminology, the register for each major character, and the rules for handling cultural references. This is not boilerplate; it is specific to the game world and it goes to every translator on the project before translation begins.

For large narrative projects, we add a consistency review pass after translation: a single editor reads the full output, flags register breaks, and proposes corrections. This pass is the difference between a localization that reads as a coherent work and one that reads as a committee document.

For launch-critical content (store descriptions, trailers, marketing copy), we treat cultural adaptation separately from translation. The question is not “what does this say in French” but “what would a French player want to hear about this game.”