Localization that doesn't break your build.
Your engine never changes. Keys stay anchored, placeholders stay intact. Only the language table swaps.
The keys never move. Only the text does.
They are already playing. And they are complaining.
"Great 2D platformer, the story is legendary — the only bad part is there's no Turkish language support."
original: "kardeşim 2d güzel bir platform oyunu hikayesi de efsane sadece türkçe dil desteği yok bir o kötü" · translated by LoreString
“Eleven years on and the author still hasn’t added Chinese. I knew this game back in primary school — I’m grown up now and I still want to finish it myself. But please, I’m begging you: add Chinese.”
“I’d recommend it if it were in Polish — as it is, I have to translate with my phone.”
“Really good story — a shame only two languages are available. Otherwise, a gem.”
“Solid game. If it had Russian it would be on fire. Still great as it is — recommended.”
“It’s fun, I do recommend buying it. … They told me I have to beat the game once before I can install the Korean file. I’ve been at it for eight hours and I’m still playing in English. I’m furious.”
“Really good game — but it has been more than ten years since Undertale came out and there is still no Spanish translation.”
Originals shown under each quote · translated by LoreString
These are the demands we read — in whatever language they are written. Different games, different countries, the same sentence underneath: I want your game, and I cannot play it in my language.
Closing that is the work. Turkish and English stay in house, start to finish; every other language goes to a translator who is a native speaker of it.
Translation is easy.
Not breaking the game is the hard part.
What actually breaks a localized build.
Mobile breaks differently. We know how.
Same job, harder traps. A string table is a string table — but on phones a wrong specifier doesn't look odd, it terminates the app.
Format specifiers
Android writes %1$s, iOS writes %@. Translate one into the other and the app doesn't render wrong — it crashes. Every specifier is counted and typed before and after.
Plurals
Android <plurals> and iOS stringsdict carry one entry per quantity rule. Turkish and English do not split those rules the same way: one direction has to collapse entries, the other has to add them. Get it wrong and the count silently disappears from the sentence.
Overflow
Across the strings we have moved between the two, Turkish runs roughly 20–30% longer than English — and that cuts both ways. Into Turkish, a label sized for a 360 dp button comes out clipped; into English, the same label leaves the button half empty. Length is measured per string against the widget it lives in. Length is measured per string against the widget it lives in.
A missing brace is a crash, not a typo.
Every line is counted before and after translation — placeholder count and order both. One of these ships. The other never leaves the pipeline.
source · Race added to the bulletin: {0}
translation · Yarış bültene eklendi: {0}
source · Week set to {0}!
target · Hafta ayarlandı.
Three steps. Your engine is never touched.
The pipeline is built once, and that is the part that takes the time. Extraction, placeholder checks and the merge back into your format are ours; the Turkish is written by a translator whose first language is Turkish. For a language outside the English–Turkish pair, the pipeline and the verification are still ours, and the translation is written by a native speaker of that language, hired for the job.
Extract
Your build is scanned, every visible string catalogued and given a permanent key. Data values are separated out. The catalogue comes back for you to review.
Translate
Native translation with a locked glossary, placeholder checks and UI-length control, so a label that runs longer or shorter than the source is caught before it breaks the button it sits in.
Merge & verify
The table is merged back in your format, then verified line by line. You get a build that runs identically — in another language.
One rate. No minimum you can't afford.
Agency translation runs $0.09–$0.18 per word; we sit at the bottom of that range. Then localization engineering lands on top as a separate line — $5,000 to $20,000 when a game wasn't built to hold translations, typically 25–35% of the whole budget. That line is the job here, not an extra.
- JSON, CSV, PO, XLSX, .strings — your call
- Locked glossary and consistent terminology
- Placeholder and UI-length verification included
- Credit in your game's credits, if you want it
- Patch and DLC updates per job or monthly
Keys anchored for life, placeholders mapped and verified, data values separated from text, your engine untouched. This is the line agencies call localization engineering. They do it by hand. We built the pipeline once — so you pay once, and not five thousand.
Already have a string table? The setup drops, often to zero. I tell you after looking at your build — free, before any quote.
A Turkish game that does not speak English yet
This is work we have already shipped: a full Turkish game taken to English before it reached Steam — around 1,900 keys, a fifth of them carrying placeholders, zero engine changes. Keys stayed where they were, placeholders were matched and checked one by one, and the build went back the same day.
Talk about itReview counts and language support read from each game's Steam store page on 4 October 2026; a studio may add Turkish at any time. Other figures: Steam Hardware & Software Survey (Sept 2026) · industry localization cost reports, 2026.
Same words. One line instead of four.
- Translation · 18,000 words$1,620
- i18n engineering$5,000+
- Linguistic QA · hourly$900
- Format delivery & glossary$400
- Translation · 18,000 words$1,620
- Pipeline setup · one time$2,000
- Placeholder & UI-length verificationincluded
- Format delivery & locked glossaryincluded
You pay for the pipeline once. After that a second language is the word rate alone — verification included, no engineering, no second setup. The translation itself is done by a native speaker of the target language, hired for the job.
Agency figures are the published market range for a game that was not built to hold translations. Yours may be lower — ask them to quote the engineering line separately and compare.
Four ways in that aren't a discount.
Budget is a real constraint for a small studio. None of these lower the quality — they change what you trade.
Two games, one setup
Shipping more than one game? The second game's pipeline setup is free — the glossary, the format adapter and your conventions are already built. You pay words only. Built for studios with something still unreleased.
Founding studio · first 5 games
Setup fee halved. In exchange: a credit line in your game's credits, permission to name the game as a reference, and two sentences about the work when it ships.
Interface first, story later
Menus, buttons, errors, tutorial — the fifth of the text that makes the game playable. The rest ships in a later pass. Same rate, smaller job.
Setup paid at launch
The setup fee waits until the game is selling. For studios with a release date and no cash before it.
Drop your own file. See what would break.
Placeholders, missing keys, plural rules, overflow risk — the same checks that run on every job. One file gives you a structure report; two files compare source against translation.
Prefer a page of its own? The checker also lives at /arac/ — same engine, shareable link.
Send 500 words. Judge for yourself.
Pick your messiest 500 words. They come back as a side-by-side report: every line, every key, every placeholder verified. You judge the quality. The build-ready file comes with the order.