Visible headlines were not enough

The first English test opened in English but still revealed Korean popup titles, inventory descriptions, errors, and stage transitions. We had translated prominent copy while missing strings created by conditional actions. In an escape game, those small lines determine the next action, so a missing translation feels like a progression bug.

We rebuilt the inventory around the player's route rather than source files: setup, intro, investigation, collection, invalid use, hints, death, stage results, story transitions, and final sharing. Every reachable line is now managed as a Korean and English pair.

Preserving puzzle rules

Puzzles tied to numbers, shapes, strokes, or directions can change when translated literally. English also expands beyond Korean button widths, and inconsistent object names can break a clue connection. A terminology list keeps important names identical across inventory, objectives, and every stage.

Atmospheric narration is separated from operational instruction. Story lines preserve ambiguity, while buttons and objectives remain concise. Repeated actions such as Next use the same label across stages, and result screens keep time, hints, and deaths in one consistent order.

URLs and language state

The official site uses /kr and /en, while the game uses the same paths on the play subdomain. Every article has a language-specific permanent URL, and the switch links to the corresponding version of the current page. Canonical and hreflang metadata identify the pair as translations rather than duplicates.

We do not force redirects from browser language. A player's explicit choice wins, and shared links open in the language chosen by the sender. Document language is set to ko or en for assistive technology and search indexing.

Full-route review

Source search cannot reveal conditional popups or clues embedded in images. We play the English game through all five stages, intentionally misuse every item, fail passwords, trigger death, request hints, and restart. Inventory is opened both immediately after collection and before use to compare titles and descriptions.

Changes reach the test deployment first. Line wrapping and target size are reviewed on desktop and mobile before the production game is updated. Quality is measured by whether both languages lead to the same action, not grammar alone; an overlong English line is rewritten instead of shrinking the interface.

The baseline for future episodes

Future episodes will not wait for Korean completion before localization begins. The terminology and English clue are drafted when each puzzle rule is locked. Conditional copy must exist in both languages, and world documents expose the same public information so one edition never reveals an unintended clue.