Translation memory is one of the most valuable assets a translation team owns. It holds every approved sentence you have ever paid for. But what happens when you change tools, work with a new agency or need to share that memory with a freelancer who uses different software? That is exactly what TMX files are for.
What Is a TMX File?
A TMX file (Translation Memory eXchange) is an XML-based file used to move translation memory data between different tools. It was created so that translation memories are not locked into one CAT tool. The current widely used version is TMX 1.4b, and the TMX 1.4b specification is published by GALA, the Globalization and Localization Association.
Almost every professional translation tool can export and import TMX, which makes it the closest thing the industry has to a universal format for translation memory.
How a TMX File Is Structured
A TMX file has two main parts:
- Header: information about the file, such as the tool that created it, the source language and segmentation type.
- Body: a list of translation units, marked
<tu>. Each unit contains two or more language variants, marked<tuv>with anxml:langattribute, and the actual text inside a<seg>element.
Translation units can also carry metadata, such as creation date, author, project or client. This metadata is useful when you want to filter or clean a memory later.
When You Need a TMX File
- Switching CAT tools or TMS platforms. Export your memory as TMX from the old tool and import it into the new one.
- Working with external vendors. Send a TMX to an agency or freelancer so they can reuse your approved translations.
- Merging memories. Combine memories from different teams, products or acquisitions into one resource.
- Backing up your data. A TMX export is a tool-independent backup of your translation assets.
- Training or customizing engines. Clean bilingual data is the starting point for customizing machine translation.
TMX vs XLIFF vs TBX
| Format | What it stores | Typical use |
|---|---|---|
| TMX | Translation memory: aligned source and target segments | Moving memories between tools |
| XLIFF | A document or set of strings being translated | Exchanging files for translation |
| TBX | Terminology: terms, definitions, usage notes | Moving glossaries and term bases |
If you want to learn more about the other two formats, read what is an XLIFF file and our guide to terminology management.
How to Keep Your TMX Files Clean
A translation memory is only as good as the data inside it. Before importing a TMX into a new tool, it pays to clean it:
- Remove duplicates and keep the most recent approved translation.
- Delete empty or untranslated units where the target is missing or identical to the source.
- Check language codes, for example that
en-USandenare used consistently. - Remove outdated terminology, such as old product names.
- Check tags, because inline formatting codes from different tools may not map perfectly.
- Keep metadata such as client or domain, so you can filter later.
Common TMX Problems
- Encoding errors that turn accented characters into strange symbols. Always use UTF-8.
- Very large files that are slow to import. Split them by language pair or client.
- Inline tags that do not match the new tool, causing formatting warnings.
- Low-quality legacy segments that pollute suggestions in the new tool.
Translation Memory in WordBeam
WordBeam includes translation memory on every plan, with exact and fuzzy matches shown directly in the CAT editor. Translation memory is applied before AI engines are called, so you never pay twice for content you already own. To learn how much this can save, read translation memory software explained, or see how WordBeam's translation memory software works in practice.
Frequently Asked Questions
How do I open a TMX file?
You can open it in any text editor because it is XML, but it is easier to read in a CAT tool or a dedicated TMX editor that shows source and target side by side.
Does a TMX file include terminology?
No. TMX stores translated segments. Terminology is exchanged in a separate format, usually TBX, CSV or XLSX.
Will my translation memory work the same after importing a TMX?
The text will, but match rates can differ slightly because each tool segments and scores text in its own way.