DEV Community

GTE LOCALIZE
GTE LOCALIZE

Posted on AI-assisted

Localizing a Mobile Game: An Eight-Step Pipeline and the Bugs It Prevents

Mobile game localization is not just swapping strings. UI, story, art, audio, store listing and legal text all have to feel native, and the work does not end at launch because live games keep shipping content. Here is a pipeline that treats localization as part of the build, with the bugs each stage is meant to catch.

What actually needs localizing

User interface, narrative text, visual assets (icons, banners, costumes), audio and voice-over, app store metadata, in-game offers, and legal and support content. Missing any one can produce bad reviews or a poor launch in that market.

The eight stages

  1. Plan. Pick target languages and launch priorities, estimate word count and voice-over volume, map the schedule to sprints, and prepare a localization kit with a style guide, glossary, screenshots and reference video.
  2. Internationalize. Support Unicode, leave extra UI room, choose fonts that cover Arabic, Cyrillic and Vietnamese, and move text out of code so strings are not hard-coded.
  3. Export source text. Export to CSV, JSON, XML, XLIFF or PO, connect to a translation management system, and keep naming consistent so updates do not conflict.
  4. Translate and localize audio. Work in batches with CAT tools, translation memory and a termbase. Adapt tone, humor and references rather than translating literally.
  5. Integrate. Import translated text and audio into the build with an automated pipeline, and run quick checks for broken tags and placeholders.
  6. Linguistic QA. Native testers play the build and check UI text, dialogue, missions and tutorials for overflow, font problems, mistranslations and tone. Mid-size titles often spend 8 to 12 weeks here.
  7. Compliance and submission. Match store and platform guidelines and official terminology.
  8. Release and update. Repeat a short translate, integrate and QA loop for patches, events and downloadable content.

Three bugs worth designing against

Strings without context. A single word such as "Play" or "Charge" can mean several things. Give strings descriptions, screenshots or video, group them by screen or feature, and use in-context testers.

Text expansion. German text typically runs 20 to 35 percent longer than English, so build flexible containers with a safety margin and use pseudo-localization early, before the first export. Chinese, Japanese and Korean strings use fewer characters but need larger font sizes and taller lines to stay legible.

Hard-coded text. Strings embedded in code or images never reach translators. Externalize them before the localization pipeline begins.

Why teams bother

Surveys cited in the source material say 76 percent of consumers prefer buying in their native language, and over 90 percent of players in Taiwan, Korea, China and Japan choose localized games over English ones. Genshin Impact, localized into 13 languages with four voice-over options, earned over $4 billion, with about 45 percent from outside China. At the other extreme, Arena of Valor, a stripped-down international version of Honor of Kings, earned about $71 million against the original's reported $15 billion.

If you are looking for the UI side in more detail, the previous post covers four UI bugs that only show up in other scripts, and the first post in the series covers five i18n bugs we keep finding in real projects.

Top comments (1)

Collapse
 
alexshev profile image
Alex Shev •

Localization is strongest when it is treated as structured data with context, constraints, and screenshots—not a bag of strings. Pseudolocalization plus automated checks for missing placeholders and text expansion catches many defects before language reviewers ever need to see them.