Video Game Localization Cost: A Pricing Guide for 2026
Localization is one of the few line items in game development you can estimate almost exactly before the work starts. Art and code are guesses until someone builds them. Translation is not. You have a word count, you have a set of target languages, and you have a per-word rate. Multiply the three and you are most of the way to a number.
The catch is everything around the translation: linguistic QA, engineering for text that grows when it changes language, voice-over, and the updates you will keep shipping after launch. Those are where budgets quietly blow out. This guide walks through the whole pricing model, gives you a table of typical market ranges, and works a real example end to end so you can put a defensible figure in your budget.
How game localization pricing works
The core of localization pricing is simple: translation is billed per source word. "Source" matters. You pay based on the word count of your original text, usually English, not the translated output. That is good news, because it means you can price a job from the script you already have, before anyone translates a thing.
A source word rate bundles the translator and, in a proper workflow, a second linguist who reviews the translation. Rates move mainly on one factor: the language pair. Common languages with deep supply of translators cost less. Rarer pairs and complex writing systems cost more. Everything else (subject matter, tone, turnaround) nudges the rate inside that band rather than setting it.
Two more billing facts to hold onto. First, minimum project fees are standard. A 300-word app-store description will not be billed at 300 times the word rate; there is a floor, often a few hundred dollars per language, because setup and coordination cost the same whether the job is small or large. Second, linguistic QA is billed by the hour, not the word, and it is a separate line from translation. More on that below.
Localization is one of the least-listed services in our directory. Of the 422 outsourcing studios we track, only 31 list localization, so supply is thinner than for art or programming and it pays to shortlist early. If you want to see who offers it, browse the outsourcing directory and filter for it.
Typical market rates by language tier
Per-word rates cluster by how common the language is and how much engineering the writing system demands. The ranges below are typical market ranges, not a proprietary survey; treat them as the band a quote should land inside, then expect movement based on volume, subject matter, and turnaround.
| Language tier | Example languages | Typical rate (per source word) |
|---|---|---|
| Common European | French, Italian, German, Spanish, Brazilian Portuguese | $0.08-0.15 |
| Nordic & Central/Eastern European | Dutch, Polish, Russian, Swedish, Turkish | $0.10-0.18 |
| CJK & complex-script | Simplified/Traditional Chinese, Japanese, Korean, Arabic | $0.12-0.20 |
A few notes on reading the table. The common European set is often sold as a bundle called EFIGS (English, French, Italian, German, Spanish); it is the default first wave for most games because supply is large and rates are lowest. The CJK tier costs more per word and also carries the heaviest engineering, because those scripts need font and glyph support that Latin text does not. The middle tier sits inside the overall $0.08-0.20 band and is where a lot of European coverage lands once you go past the big five.
The chart makes the most important budgeting point for you: the number of languages is a straight multiplier. Every language you add pays the full translation and QA bill again. This is why choosing languages by expected revenue, rather than translating everything because it feels complete, is the single biggest call you make on cost.
What drives the cost
Six factors set your final number. In rough order of how much they move it:
Word count. The base multiplier. A 5,000-word puzzle game and a 120,000-word RPG are different projects at the same rate. Cut filler in English before you localize; you pay for every word in every language.
Number of languages. The other multiplier, shown above. Each language is a fresh translation, review, and QA pass.
Language pairs. Which languages, not just how many. Eight common European languages cost less than eight that include Japanese, Korean, and Arabic.
Linguistic QA. LQA is a native speaker playing the actual build and checking that strings read well in context, fit their UI boxes, and are not clipped or mistranslated by context they could not see in a spreadsheet. It is billed hourly, in the QA rate band of roughly $25 to $80 per hour, with experienced native testers toward the upper half. Budget real hours here; skipping LQA is the classic false economy that ships "press START to begin" as a broken string in five languages.
Voice-over versus text-only. Text is per word. Voice is priced by recorded line or studio hour, plus talent fees, and for a dialogue-heavy game it can cost more than the entire text translation. Voice is where "let's just do everything" budgets explode.
Localization engineering. The plumbing. Pulling strings out of code, handling text expansion (German and Russian often run longer than English and overflow buttons sized for English), adding fonts and glyph support for CJK, and wiring up right-to-left layout for Arabic. If the game was not built to be localized, this line grows fast. If strings live in external files from day one, it nearly disappears.
And one that is easy to forget at kickoff: ongoing updates. A live game keeps generating text. Every patch, event, and new character adds words that need the same treatment. Localization is a recurring cost for a live title, not a one-off at launch. Price it as a retainer or per-word top-up, and keep the same linguists so voice and terminology stay consistent.
A worked example: a 40,000-word RPG
Numbers make this concrete. Say you have an RPG with a 40,000-word script and you want eight languages: five common European (French, Italian, German, Spanish, Brazilian Portuguese) plus three Asian (Japanese, Korean, Simplified Chinese). Text-only, no voice yet.
Translation is word count times rate times languages. LQA is hours per language times an hourly rate. Here is the build:
| Group | Languages | Words each | Rate/word | Translation | LQA (40 hr × $45) |
|---|---|---|---|---|---|
| Common European | 5 | 40,000 | $0.12 | $24,000 | $9,000 |
| CJK | 3 | 40,000 | $0.16 | $19,200 | $5,400 |
| Total | 8 | $43,200 | $14,400 |
So the text-only base is about $57,600: $43,200 to translate and review, plus $14,400 for eight LQA passes at 40 hours each. That LQA estimate assumes a tester needs roughly a working week per language to play through a 40,000-word RPG and check strings in context; a linear game with less branching needs fewer hours, a sprawling one with lots of conditional text needs more.
Now the lines the base does not include:
- Localization engineering. If your build is loc-ready, this is small. If strings are baked into art and code, or you need CJK fonts and Arabic layout added, budget $5,000 to $20,000. Our example skips Arabic, but the three Asian languages still need font and glyph work.
- Minimum fees. On a 40,000-word job they do not bite, but on a small game they can dominate the bill.
- Voice-over. Out of scope for this example on purpose. If you added full voice in even two of these languages, you could roughly double the total. This is why most teams ship text in many languages and voice in few.
Put together, a text-only, loc-ready version of this project lands somewhere around $60,000 to $75,000 once you fold in engineering. That is a real, traceable figure you can defend in a budget meeting. For a broader breakdown of how localization sits next to art, programming, and audio in a full project budget, see our complete guide to game outsourcing costs.
Text-only versus full voice-over
The single biggest fork in a localization budget is whether you record voice. Text-only localization means every language reads the game in its own words but hears the same original voice acting, if any. Full localization records new voice in each language.
Text-only is the default, and for good reason. It gets your game understood everywhere for a predictable per-word cost. Players in most markets accept subtitled or text games without complaint. Voice-over is expensive, slow, and hard to change: if you patch a line of dialogue after recording, you are back in a studio with the same actor, assuming they are still available.
The practical rule most teams follow: localize text into every language where you expect meaningful sales, and add voice only in the two or three markets that will pay for it. Japanese and Chinese voice, for instance, can move the needle in those markets in a way that justifies the spend; a fourth or fifth voice language often does not.
If you are recording voice, get the script locked first. Voice recorded against a script that then changes is money spent twice.
Localization engineering and LQA, in plain terms
These two lines are where inexperienced teams underspend and pay for it later.
Localization engineering is not translation; it is making the game able to hold translations. The usual failures are avoidable and predictable. Text baked into image files has to be re-rendered per language. Strings hardcoded in source have to be dug out one by one. UI built to fit English clips German and Russian, which run longer. CJK and Arabic need fonts and, for Arabic, right-to-left layout. Fix these at the architecture stage and the cost is near zero. Retrofit them after the game is built and you are paying a programmer to unpick decisions that were free to get right the first time.
LQA is the safety net. A translator working from a spreadsheet cannot see that a string appears on a button, or after a character's name, or only when the player is low on health. The LQA tester sees it in the running game and catches the context errors no spreadsheet review can. This is skilled work by native speakers, billed hourly in the QA band, and it is the difference between a game that reads as localized and one that reads as machine-translated. Do not cut it to save a few thousand dollars on a project that cost tens of thousands to translate.
Machine translation and where it fits
You will be offered a cheaper tier called MTPE, machine translation post-editing. A machine translates first, then a human editor cleans it up. It is billed per word at a discount to full human translation, though the size of the discount varies by vendor and by how much editing the output needs.
MTPE is a real cost-saver for the right content: menus, settings, store descriptions, patch notes, and other functional text where the phrasing is plain and the risk of a howler is low. It is a poor fit for anything with voice: character dialogue, jokes, wordplay, lore. Machine output flattens tone and misses context, and a light post-edit will not rescue a joke it did not understand. A common approach is to split the script, run functional text through MTPE and give narrative text full human translation, so you pay the premium only where it earns its keep.
Be wary of a vendor selling raw machine translation with no human pass as "localization." It is not, and players will spot it in the first menu.
What a localization quote should include
A quote that is only a per-word number is a quote that will grow later. A complete one lists these lines so nothing surprises you:
- Translation (per source word, per language) and whether a second-linguist review is included.
- LQA hours per language and the hourly rate.
- Localization engineering: string extraction, text expansion handling, fonts, and layout.
- Minimum fees per language, if any.
- Voice-over, quoted separately by language if you are recording it.
- Updates: the rate for post-launch content, and whether the same linguists are retained.
If a line is missing, ask before you sign. Vague quotes are where the overruns hide.
How to keep the bill sensible
You control most of the cost before a single word is translated:
- Trim the English first. Fewer source words, paid in every language. This is the cheapest saving available.
- Build loc-ready. Strings in external files, never in art or code. This deletes most engineering cost.
- Tier your languages. Launch EFIGS, add CJK when the data supports it, add the long tail later. You do not have to buy all languages on day one.
- Reuse and reduce. A translation memory means repeated strings are translated once. Ask whether your vendor uses one; it should.
- Keep the same team for updates. Continuity keeps terminology and voice consistent and avoids re-briefing on every patch.
- Voice only where it pays. Text everywhere, voice in the few markets that earn it back.
Localization is worth doing well because the alternative shows up in reviews. A broken or clumsy translation reads as a broken game to the players living in it, and no amount of good art hides it. But "well" does not mean "everything, everywhere." It means a tight source script, a loc-ready build, languages chosen by revenue, and real LQA on each one.
If you want to line up quotes, our roundup of the best game outsourcing companies is a faster start than a cold search, and you can browse the full directory to filter for studios that list localization and contact them directly with no platform fees.
Ready to localize? Browse our directory of 420+ verified game development studios, filter for localization, and contact them directly.