How to convert a Trados .sdltm memory to TMX without Trados

What an .sdltm file is

An .sdltm file is a Trados Studio translation memory. Inside, it is an SQLite database. Its main table holds the translation units: for each one, the source and target segment, the date it was made and last changed, and the users who did it. A second table names the memory and its two languages. Tags sit inside each segment, so a sentence keeps its formatting marks wherever it goes.

You may have one because an agency sent it with a job, because it came inside a Trados project package, or because it holds your own years of work in Trados. Either way, a tool that doesn’t read .sdltm can’t use it until it is converted.

What TMX is

TMX, Translation Memory eXchange, is an XML format “for the exchange of translation memory (TM) data between computer-aided translation” tools. It was first released in 1997, and its current version, 1.4b, dates from 2005 (Wikipedia, checked 20 October 2026). Each unit is a <tu> element with one <tuv> per language. Because it is plain text, you can open a TMX file in any text editor and read it.

Converting it in Lexora Studio

1. Install Lexora Studio. The 14-day free trial includes every feature and needs no card, so a one-off conversion costs nothing.
2. Open a project. Lexora reads memory files for the open project’s languages, so open or make a project in the memory’s languages. If the languages differ, Lexora asks whether to import the memory with its own languages or to swap source and target.
3. Make an empty memory. Open Memories and glossaries in the sidebar and choose New memory.
4. Import the .sdltm. Use the import button on the memory’s row and pick the file. A large memory imports in the background, with the number of units read so far and a cancel button. When it finishes, Lexora says how many entries came in and how many were skipped.
5. Export it as TMX. On the same row, choose the TMX export. Lexora writes a TMX 1.4 file named after the memory.

Moving from Trados for good? Switch from Trados or memoQ, on the Welcome page and in Settings, finds the memories in your usual Trados folders and imports them in one step. The help article on switching shows how.

What comes across, and what stays behind

Lexora reads these from each unit and writes them into the TMX:

• the source and target text;
• the tags, in their places: a start and end tag that belong together become a pair in TMX, any other tag a placeholder;
• the two languages as Trados names them, such as en-US and de-DE;
• the date the unit was made and the date it was last changed;
• the user who made the unit, and the name of the memory.

These stay behind:

• units with an empty source or target, and units that repeat the same source and translation, which become one entry; Lexora counts both as skipped;
• the user who last changed a unit, when that differs from the one who made it;
• anything else the file keeps, such as how often a unit was used, or fields an agency may have set up. Lexora reads only what is listed above.

If your agency relies on those fields, for example to filter matches by client or project, ask it how it wants the memory before you convert it.

Checking the result

Before you rely on the TMX, open it in a text editor and look at a few units. Then import it into the tool you are moving to and translate a sentence you know is in the memory: the match should appear with its tags. Keep the original .sdltm; converting doesn’t change it.

What we tested, and what we haven’t

Our automated tests build memories in the table layout Trados Studio uses, with paired and single tags, dates, users, an empty unit and a duplicate. They import each memory, match a sentence with tags against it, export the entries to TMX and read the TMX back unchanged. Another imports a memory of 100,000 units, reporting progress as it goes, and checks that a cancelled import adds nothing. We ran them on 20 October 2026, on Lexora 0.8.5, and they passed.

What we haven’t done for this post: converted an .sdltm made by Trados Studio 2026, the current version, or opened our TMX in Trados or memoQ. We can’t run either here. If a memory doesn’t convert as described, write to hello@lexorastudio.com and we will correct this post in the open.

What other tools do better

The tool that made the memory knows everything in it. If you or a colleague have Trados Studio, its own export may keep the fields and details that Lexora leaves behind, so it is worth asking before you convert a memory full of them. Lexora also runs only on Windows 10 and 11.

Read next

• Translation memory software for freelance translators: how Lexora keeps a memory for each client
• What a translation memory is and how to start using one
• Working on a Trados package when you don’t own Trados