Free online
translation memory cleaner
Remove duplicate translation units, junk segments and machine translation from your TMX file. Preview every change before anything is cut and see exactly how much smaller it gets, then download the cleaned TM.
◇ EU-hosted · GDPR-compliant · Uploads purged the moment cleaning ends · Zero trackers
━━ what TM Cleaner removes
duplicate translation units
Every variant of a segment competes for the same match. TM Cleaner keeps one per group (the most recent, the most used, or your trusted translator's) and you can pick a different winner for any group.
junk segments
Units that are only URLs, email addresses, numbers, inline tags, placeholders, file paths or punctuation. They inflate word counts and add noise to fuzzy matches without ever helping a translator.
machine translation
Units whose metadata shows a machine-translation origin (Google, DeepL, Bing, Yandex, or patterns you add), so raw MT stops passing as human-confirmed leverage.
broken and untranslated units
Empty and untranslated targets go by default. Optional checks catch placeholder and tag mismatches, encoding damage (mojibake, which can also be repaired in place) and length outliers.
Every rule shows how many units it would remove before you commit, and each one can be switched off. See every rule and preset.
━━ how to clean a translation memory
- 01
export your TM as TMX
In Trados Studio, memoQ, Phrase, XTM, Smartcat, OmegaT or any other CAT tool, export the translation memory to a .tmx file.
- 02
upload and preview
Drop the file in and review exactly what would be removed, group by group. No account needed for files up to 50 MB. Change any rule and the numbers update.
- 03
clean, download, re-import
Clean and download, no account needed for files up to 50 MB. Get the cleaned TMX, plus removed.tmx and removed.csv so you can check every cut, and import it back into your CAT tool like any TMX. Sign in with an emailed link (free, no password) for files up to 2 GB.
What to remove, what to keep and how to check the result: the full guide to cleaning a translation memory.
━━ what is a TMX file?
A translation memory (TM) is the database of source and target segment pairs, called translation units, that your CAT tool reuses when the same or a similar sentence comes up again. TMX (Translation Memory eXchange) is the open XML format every major CAT tool uses to export and import that database.
So TM Cleaner works on the TMX file: export your translation memory as TMX, clean it here, and import the cleaned file back. You don't need the CAT tool that created it, and the output keeps the same TMX structure, so any tool can read it.
━━ clean a TM from your CAT tool
Any tool that exports TMX works. These pages cover each tool's export steps and the junk its exports typically carry:
Or start from the job you need done:
━━ frequently asked
Is TM Cleaner really free?+
Yes. There is no payment and no card. You can preview, clean and download a file up to 50 MB without an account. Sign in with a one-time link sent to your email (no password) to clean any number of files up to 2 GB each.
What happens to my translation memory after I upload it?+
It is deleted from storage the moment cleaning finishes, unless you choose to keep it for 24 hours so you can clean it again with other settings. The download links work for 1 hour, then the cleaned files are deleted too.
Will the cleaned file import back into Trados, memoQ or Phrase?+
Yes. The cleaned file keeps your TMX's own header and version, so you import it as a new translation memory or into an existing one, exactly as you would any TMX file. If your CAT tool refused the original (an invalid character, a header that says *all*, one language spelled three ways), TM Cleaner repairs the file and its import check fixes the codes.
Can it read Excel, CSV or several files at once?+
Yes. Drop several TMX files, or a .zip of them, and they become one TM with duplicates removed across all of them. A bilingual .xlsx, .csv or .tsv becomes a TMX, with the languages read from its first row.
Can I see what will be removed before anything changes?+
Yes. The preview lists every duplicate group with its winner and every junk category with examples. You can switch rules off, keep individual units or choose a different winner, and nothing is removed until you confirm. Your original file is never modified, and removed.tmx holds every unit that was cut, so you can re-import anything you want back.
How large a TMX file can it handle?+
Up to 2 GB per file. The engine streams the file instead of loading it into memory, so TMs with millions of units work; a 1 GB file typically cleans in under ten minutes.
Which languages does it support?+
Any language pair a TMX file can hold, including CJK, Indic and Thai. Text is Unicode-normalised before comparison, so duplicate detection and length checks are fair across scripts.