The Aligner pairs your original documents with their existing translations, sentence by sentence, and gives you a single file of aligned pairs. Use it to turn translations you already have (from before wxrks, or from another tool) into a Translation Memory (TM), a bilingual spreadsheet, or a parallel corpus (a collection of texts paired with their translations, used for example to train or evaluate AI models).

Who is this for? Account Admins and Project Managers. The Aligner appears only for these two roles. Clients and Vendors don't see it in the menu.

What you'll need:

In the sidebar, click Tools, then The Aligner.

ℹ️ Note: Only the accepted file types listed in the panel can be added. If you select other files, The Aligner rejects them with Not accepted: {file names} and adds nothing. Text files (.txt, .md, .json, .html, .htm, .xml) must be saved with UTF-8 encoding.

💡 Tip: Detection selects the base language, such as Portuguese, and the output file is tagged with that code (pt). If the TM you'll import into uses a regional variant, select it instead, for example Portuguese (Brazil), so the file is tagged pt-BR. Do the same for the source language.

Within each batch, The Aligner first pairs source and target files by file name. It ignores upper and lower case, the file extension, and a language code at the end of the name. For example, all of these pair with each other:

Files left without a name match are then combined, on each side, and aligned as one block of text.

💡 Tip: Give each translation the same name as its original, plus a language code. Name-matched pairs align document to document, which gives more accurate results than combined, unmatched text.

⚠️ Warning: If, after name matching, only one side of a batch still has files left over (for example, an extra target file with no source to pair with), that content is skipped. A warning in the results tells you how many documents were skipped.

Click Settings in the top-right corner of the page.

Click Save. These settings, and the output format, are saved in your browser. They're still there the next time you open The Aligner on the same browser.

ℹ️ Note: The Aligner uses its own built-in sentence rules. Custom Segmentation Rules configured in your account don't apply here.

The threshold is a trade-off: raise it to keep only pairs the AI model is sure about, lower it to keep more pairs and review them yourself.

Example: a run produces 120 aligned units. 95 score 0.80 or higher, 18 score between 0.50 and 0.79, and 7 score 0.20.

⚠️ Warning: Avoid setting the threshold to 0.20 or lower. When the AI model can't align part of a document, The Aligner falls back to pairing sentences in order (first with first, second with second) and gives those pairs a confidence of 0.20. These pairs are often wrong. At the default 0.80 they're discarded. At 0.20 or lower they end up in your output, and wrong pairs in a TM become wrong matches in future projects.

To cancel, click Stop. The run ends and the page shows Alignment was stopped.

When the run finishes, the Output panel shows Alignment complete: {N} units ({format}) and a summary:

ℹ️ Note: Units are counted per language. Ten source sentences aligned into Portuguese and French give Alignment complete: 20 units (TMX).

Click Download to save the file. It's named after the run's format, for example alignment.tmx. All batches go into one file: each source sentence appears once, with its translation in every target language you aligned.

ℹ️ Note: The downloaded file always uses the format of the run, even if you change Format afterwards. To get a different format, choose it and click Run Alignment again.

⚠️ Warning: Download the file before you leave or reload the page. The result isn't saved to your account, and once you leave the page the Download button is gone.

Example: you aligned 10 English sentences into Portuguese and French, and you import the file into an English (United States) → Portuguese (Brazil) TM.

⚠️ Warning: If the TM should keep only its own language pair, choose SELECTED before you upload. With ALL, every extra language in the file becomes part of the TM.

For more on importing, see Import in All about Translation Memories.