Anonymize a translation memory before sharing it

Sending a TM to a new vendor, a client or an MT provider shares more than translations: every unit names who created and edited it, notes carry internal comments, and the text itself holds email addresses, phone numbers and customer names. GDPR says you should know what you're handing over.

do it now →see a live demo

━━ the problem

every unit names a translator

creationid and changeid record who wrote and who edited each unit; props and notes carry project codes and internal remarks. That's personal data about your team, in every line.

the text is full of contact details

Support emails, phone numbers and bank details end up in segments from contracts, signatures and invoices — and from there in the TM.

client names you can't share

A TM built for one client mentions them, their products and their people. Find-and-replace in a 2 GB file isn't a plan.

━━ how TM Cleaner handles it

Pick “Share safely”: translator names, usage counts, properties and notes are removed, and emails, phone numbers, IBANs and card numbers in the text become [EMAIL], [PHONE], [IBAN] and [CARD]. Add names to hide and they become [NAME]. Tags inside segments are never touched. The result says what was masked, by kind.

✓masks the same way in both languages, so the pairs still line up
✓phone numbers are masked when they're written like phone numbers; dates and product codes stay
✓card numbers are checked (Luhn) before they're masked
✓removed.tmx keeps the originals for you; share only the cleaned file

━━ frequently asked

is my TM stored while you anonymize it?+

only while it's processed: the upload is deleted the moment cleaning ends, and the results an hour later. see the privacy page.

does masking change the translation memory's matches?+

a masked segment matches other segments with the same placeholder. for MT or LLM training that's what you want; for a TM you'll keep using yourself, share the masked copy and keep the original.

which names are masked?+

the ones you list (whole words, any case). there's no automatic name detection: it would mask product names and miss real ones.

━━ related pages

Translation memory cleaner for Trados Translation memory cleaner for memoQ Remove duplicate translation units from a TM Translation memory cleaner for Phrase (Memsource)Clean a TMX for MT training data Clean a translation memory for AI translation Translation memory cleaner for XTM Cloud Remove machine translation from a translation memory Translation memory cleaner for OmegaT Translation memory cleaner for Smartcat Translation memory cleaner for Crowdin Fix a TMX that won't import Merge TMX files into one translation memory Convert Excel or CSV to TMX Reverse a TMX or extract one language pair Remove a bad import from a translation memory guide: how to clean a translation memory

clean your tmx now →

free · no account up to 50 MB · files up to 2 GB · try the demo