Our free resources list for videogame localization covers where to read: the platform-holder documentation, the glossaries, the communities. This article covers the other half, the part that decides whether a Friday delivery happens: what you actually run. The files you will be sent, the free tools that handle them, the engine-side tooling that generates them, and the checks that catch a string before a player does.
Everything here is either free or open source, and every link was opened while writing. Where a tool is only partly free, or where a paid product is named as the alternative, it says so.
One thing has changed underneath all of it. Most of what these tools do is mechanical work with a clear input and a clear output, and mechanical work is what a language model, or a short script you write yourself, now does as well. Section 5 runs that example for real: one export, one language model pass, one script written for the job, checked end to end.
Six stages. The free tool each stage of this article relies on, and the half a language model or a script you write now takes on.
In This Article
- 1.What a Game Localization Toolkit Needs
- 2.The File Formats You Will Actually Receive
- 3.Free Tools for Extraction and Round-Trips
- 4.Free CAT and TMS Options for Game Strings
- 5.What AI Does Now, and How to Build Your Own
- 6.Engine-Side Free Tooling: Unity, Godot, Unreal
- 7.Free QA, Pseudolocalisation and Consistency Checks
- 8.Free Documentation From Platform Holders
- 9.Where to Learn Free
- 10.A Toolkit by Team Size
- 11.From Game Strings to Audiovisual Work
- Frequently Asked Questions
1. What a Game Localization Toolkit Needs
Game localization has a tooling problem that film subtitling and software documentation do not have in the same degree: the text lives inside the build. A string is not a paragraph in a document, it is a row in a table that the engine reads at runtime, and it carries variables, tags, character limits and rules that make it behave in ways a plain text file cannot.
That shapes the toolkit. You need something that can round-trip a format without damaging the tags, something that can hold a glossary and a TM across a long project, something that can count characters because a UI string has a box to fit into, and something that can verify against the game rather than only against the source document. Four needs, four groups of tools, and a fifth group that is free and regularly forgotten: the format checkers that stop a broken export from reaching the build.
Fundamentos de la localización de videojuegos
The Spanish-language edition of the foundations course.
2. The File Formats You Will Actually Receive
Before choosing a tool, know what will be dropped in your inbox. Game localisation is format-rich and every format has a failure mode.
| Format | Where it comes from | What breaks in it |
|---|---|---|
| CSV and TSV | The most common studio export: one row per string, one column per language | Encoding, embedded commas and newlines, columns that shift when a language is added |
| XLSX | Producer and marketing teams who like spreadsheets | Character limits written as free text, merged cells, comments that never reach the CAT tool |
| JSON | Web, mobile and live-service titles | Nested keys, duplicated keys, escaping and unicode in the wrong place |
| PO and POT (gettext) | Open-source and indie projects, and the GNU gettext standard | Plural forms, fuzzy matches left in the file, contexts that decide the translation |
| XLIFF 1.2 and 2.1 | Any pipeline with a CAT tool in the middle; the standard is public at OASIS | Inline tags, state attributes, notes that tools read differently |
| Unity and Unreal assets | The engine itself, exported through its own localisation system | String tables split by scene, keys with no context, missing assets on a partial build |
| Subtitle and caption files | Cutscenes, trailers and voice-over scripts | Frame rates, reading speed, and dialogue that has to match a performance |
A practical rule: never edit the file you received in place. Convert in, translate, convert out, and diff the result against the original structure. A single lost placeholder in a JSON key can crash a screen.
3. Free Tools for Extraction and Round-Trips
3.1. The Translate Toolkit
The Translate Toolkit, developed in the open at translate/translate, is the free format engine of this field: it reads and writes PO, TMX, XLIFF, CSV, JSON, Android and iOS resource files, subtitle formats and more, and turns any of them into any other. When a studio sends a CSV that has to become an XLIFF for a CAT tool, or a PO that has to go back as a clean CSV, this is the tool that does it without a licence.
3.2. The Okapi Framework
Okapi is a free, open-source set of tools for exactly the same problem, oriented to pipelines rather than to one-off conversions: filters for dozens of file types, segmentation, XLIFF round-trips, a TM engine and the CheckMate QA module. Its filter configuration is what makes it valuable on game files, where a CSV has quoted fields, embedded line breaks and a column of comments that must survive the round trip.
3.3. Regex, and the discipline of a scratch file
Half of game localisation cleaning is pattern work: find the strings that contain a placeholder, the ones that are pure codes, the ones with a trailing space. regex101 is free and in a browser, it explains a pattern as you build it, and it is the safest way to test a search-and-replace before running it on a file with eight thousand rows.
One habit that saves the most time: keep a scratch copy of the original export and run a structural diff after every conversion. Comparing the number of rows, the number of placeholders and the number of non-empty targets catches almost every conversion accident before the client does.
The whole chain these converters feed is what our Videogame Localization Course teaches end to end: the formats, the round trips, the terminology and the delivery.
The Videogame Localization Course
The full workflow: files, tools, terminology and delivery.
4. Free CAT and TMS Options for Game Strings
You do not need a commercial TMS to translate a game, but you do need a tool that keeps the format intact, holds a termbase and counts characters.
OmegaT is the long-standing open-source desktop option, strong on reusable memories, glossaries and its own scriptable filters, and we compare it with the other free tools in Best Free CAT Tools in 2026. Poedit is the free and familiar editor for gettext catalogues, which is the format a large part of the indie and open-source world ships in. Weblate is the free, open-source continuous localisation platform: self-hosted it costs nothing, and for libre projects it also runs hosted, with tight integration to version control, which is exactly what a live-service title needs.
The commercial platforms are worth naming so you can answer when a client asks: Crowdin, Lokalise, Smartcat, POEditor, Transifex, memoQ and Trados. Every one of them has an entry tier and a change history on the pricing page, so check the current terms on the day rather than trusting a comparison table from last year.
If you work into European Portuguese, the same ground is taught in pt-PT in the Portuguese edition of the Videogame Localization Course, Localização de videojogos: de casual a pro.
Localização de videojogos: de casual a pro
The Portuguese edition of our videogame localization course, taught in pt-PT.
5. What AI Does Now, and How to Build Your Own
Half of this toolkit is mechanical. Converting a format, normalising a header, checking that a placeholder survived, flagging the string that goes past its character limit: each of those has a clear input and a checkable output, and each of them is now also something a language model, or a short script you write yourself, does well. That does not retire the Translate Toolkit or Okapi, and it does not change what a proper round trip looks like. It changes who you have to ask for permission.
The practical effect is that you stop being blocked. When a studio sends an XLIFF your CAT tool will not open, or a CSV whose columns shift under you, the fix no longer depends on a licence, a vendor forum or a plugin that stopped being maintained. It depends on whether you can describe the transformation precisely enough for a tool to repeat it, and that is a skill that sits next to the terminology and the editorial judgement a game localizer already has.
The other half of the same idea is that you can build the tool. A checker that runs on every delivery takes an afternoon to write and then works for years. Here is one, run for this article inside OpenClaw: a sample export, a language model pass, and the check written for the job.
key,source,limit
MENU_PLAY,Play,12
INV_POTION_USE,Use {item} on {target},40
QUEST_COMPLETE,Quest complete! You earned <b>{gold}</b> gold.,60
SETTINGS_LANG,Change the language of the interface in Settings,55
TUTORIAL_MOVE,Move with {up} {down} and jump with {jump},45
HUD_AMMO,Ammo: {current}/{max},20
Six strings from a sample build, with the character limit the studio sent in the third column. We asked a language model to translate the source column into European Portuguese, keeping every placeholder exactly as written, every tag, and every limit. This is what came back, unchanged:
key,target
MENU_PLAY,Jogar
INV_POTION_USE,Usar {item} em {target}
QUEST_COMPLETE,Tarefa concluída! Ganhaste <b>{gold}</b> de ouro.
SETTINGS_LANG,Altera o idioma da interface nas Definições
TUTORIAL_MOVE,Move-te com {up} {down} e salta com {jump}
HUD_AMMO,Munições: {current}/{max}
Then the check, which is the part you keep. This script was written for the article and now runs on any delivery in the same shape:
source rows: 6 | target rows: 6 | checked: 6 RESULT: 1 finding(s) OVER LIMIT HUD_AMMO 25 > 20
One string came back over its limit: the model translated Ammo: {current}/{max} correctly and then fitted it into more characters than the box has room for. The model did the language, the script caught the arithmetic. Keep the script.
Two things to hold on to. The first is that a model answers the question you ask and no more: nothing in that pass would have stopped a translation that ignored the studio's glossary, and a placeholder disappears the moment the instruction to keep it is missing. The second is that client strings are client material. Sending them to an external service is a contract question before it is a technical one, so check what your agreement allows before you paste a build into anything.
None of this replaces the round-trip discipline of the section above. It moves the work: the model for the language, a script for the mechanical checks, and the same script on every delivery from then on.
6. Engine-Side Free Tooling: Unity, Godot, Unreal
The strings come from the engine, so the engine's own localisation system is part of your toolkit whether you translate or manage the project.
Unity ships a free localisation package that manages string tables, locales and asset tables, and its documentation is the neutral reference for how a modern engine expects localised content to be structured. Godot handles localisation through its own system of translation files and CSV import, documented at Internationalizing games, and its CSV workflow is one of the fastest ways to see the export-and-reimport loop in practice. Unreal Engine gathers localisation targets, cultures and string tables through its own dashboard, covered at Localization in Unreal Engine.
One format detail matters across all three: the ICU message format is how plural rules, gender and select cases travel inside a string. A translator who understands a plural block will not flatten it into one form, and a reviewer who knows how to read it will not accept a string that lost its cases.
7. Free QA, Pseudolocalisation and Consistency Checks
The strongest free check in game localisation is not a spell-checker, it is pseudolocalisation: replacing every source string with an accented, expanded version of itself before the real translation starts. It surfaces text that never made it into the string table, boxes that cannot hold the expansion, strings concatenated in code and fonts missing the accented glyphs. Both Unity and Godot document the approach in the pages above.
Xbench is free and covers the mechanical side across a whole set of files: completeness against the source export, consistency of a repeated string, numbers and tags that do not match, and key-term adherence to the glossary. Okapi CheckMate does a comparable battery inside the free Okapi Framework and is easier to script for a batch. Both of them read the converted files rather than the DOCX someone made by hand.
| Check | What it catches in a game build | Free tool |
|---|---|---|
| Pseudolocalisation | Strings outside the table, boxes too small, fonts missing accents, text built in code | Unity and Godot localisation systems |
| Placeholder and tag integrity | A lost variable, a broken rich-text tag, a string that will crash a screen | Xbench, Okapi CheckMate, or your own script |
| Consistency | The same UI term translated two ways, an item name that changes between screens | Xbench, the project glossary |
| Completeness | Strings left in English, rows that never came back from the vendor | Xbench, a structural diff of the export |
| Character limits | A menu option that overflows, a tooltip that gets truncated | Limits set by the studio and checked in the build during LQA |
Game Loc in the Arabic Speaking World
A market-specific edition: culture, RTL and audience expectations.
8. Free Documentation From Platform Holders
The platform-holder requirements are free to read, and they are the only authority on what your deliverable must look like: the Steamworks localisation documentation, the PlayStation Ultimate Gaming Glossary, the Netflix Games internationalisation requirements, the Microsoft GDK localisation overview, the Epic Games Store localisation documentation and the Nintendo network service guidelines.
We listed them in detail, with what each one is useful for, in 10 Free Resources for Videogame Localization, so this article does not repeat the annotations. The professional community side lives at the IGDA Localization SIG, and the industry calendar has two events worth following for tool and workflow talks: GDC and Game Global.
9. Where to Learn Free
The toolkit above is free, and so is the reading list around it. Two of our own free resources are the fastest way in: the blog guides below walk through the field end to end, and the courses are the structured version.
- Gaming Localization in 2026: How to Start and Succeed in Video Game Translation: the entry route into the field.
- Common Problems in Video Game Localization and How to Overcome Them: the recurring defects and what a professional does about each.
- 10 Free Resources for Videogame Localization: the documentation, glossaries and channels this article points at.
The full catalogue of game and audiovisual localisation courses is at TranslaStars courses, and everything we give away sits at Free Translation Courses and Resources.
10. A Toolkit by Team Size
| Team | Free setup that works | When to pay |
|---|---|---|
| Solo translator | Translate Toolkit or Okapi for conversions, OmegaT or Poedit to translate, Xbench to check | A client-mandated CAT tool or an LQA platform |
| Small vendor team | Weblate self-hosted for the strings, shared glossaries, a structural diff script | Scale, connectors, or an enterprise client requirement |
| Studio in-house | Engine localisation system plus pseudolocalisation, version control, string tables reviewed per feature | A TMS with an API when the content team stops being one person |
| Live-service title | Weblate or a free entry tier of a commercial TMS for continuous strings | Automation and connectors, because the release cadence is the problem, not the format |
Expert on Game Localization
The advanced track for specialists who already translate games.
11. From Game Strings to Audiovisual Work
One more door. A large part of what this toolkit handles is not exclusive to games: cutscene subtitles, trailer voice-over and dialogue timed to a performance are the same objects an audiovisual translator works with, and the subtitle and caption row in the formats table above is exactly where the two fields meet.
Our Master in AI, PM & Translation for AVT is the programme that covers that ground end to end: subtitling, dubbing, media accessibility and the project management around them. Every course in the field is collected at Audiovisual Translation Courses.
Frequently Asked Questions
11.1. Can I do game localization with only free tools?
For the format work, the translation environment and the mechanical checks, yes: Translate Toolkit, Okapi, OmegaT, Poedit, Weblate and Xbench cover the whole chain. The two places where free stops are the engine build (you need access to a playable build for LQA) and any client that mandates its own TMS or LQA platform.
11.2. Which format should I ask the studio for?
Ask for what the build consumes, then convert for your own work. If you have the choice, XLIFF keeps context and tags best; CSV is the most portable and the easiest to damage. Never ask for a hand-made spreadsheet when the studio already has a string table export.
11.3. Why is my export losing placeholders?
Almost always because the file was opened in a spreadsheet and saved back, or because the conversion tool was not told that placeholders and tags have to be treated as inline codes. Use Translate Toolkit or Okapi filters, and run a structural diff on the placeholders after every conversion. A script of your own catches the same thing on the way back.
11.4. Is pseudolocalisation worth the setup time?
It is the cheapest defect detector in the whole toolkit. Twice the string length, accents in every character and a marker around every string, and you find the untranslated text, the truncated boxes and the hard-coded concatenations before a translator touches the game rather than after the first localised build fails review.
11.5. Can a language model replace these tools?
For part of the chain, yes: a first-pass translation, a format conversion, a check that runs over the delivery. What it does not replace is the decision. A model will not know that your client's glossary forbids a term, that a string is a legal notice, or that the box only fits twenty characters, unless you tell it and then check. Treat it like any other tool in this list: it does the mechanical part, and you own the result.
11.6. Do I need the engine to work as a game localizer?
You need access to a build for LQA and you need to understand how the strings are structured, but you do not need to be the engineer. Reading the localisation pages of Unity, Godot and Unreal is enough to speak the same language as the team that exports your files, and it is the difference between being sent a CSV and being asked what to send.
More on this blog: 10 Free Resources for Videogame Localization for the reading list, and Best Free CAT Tools in 2026 for the translation environment that goes with this toolkit.
