Free localization file checker

Why check every row instead of sampling

Sampling assumes quality is roughly even across a file. That holds up reasonably well for one human translator working start to finish, and much less well for machine or AI output, where one row can be perfect and the next can drop a variable. A clean sample no longer certifies the rest of the file.

The failures this tool looks for are mechanical: a placeholder either matches or it does not, a tag pair is either balanced or it is not. Those checks are cheap enough to run on every row, which leaves human review free to focus on meaning and tone — the part a script cannot do.

Your file stays on your machine

The rule engine is compiled into this page and runs in your browser. The file you pick is read locally and never sent anywhere — no upload, no server, no storage. You can confirm it: disconnect from the network after this page loads and the checker still works.

That matters if your text belongs to an unannounced title, where sending strings to a third-party service is usually the thing legal stops first.

What the file should look like

A CSV with a header row and at least three columns: a key, the source text, and the translated text. Column names are detected automatically and you can correct them afterwards. Extra columns are ignored, so an export straight out of a spreadsheet usually works as-is.