Translation memory (TM)

Translation memory (TM)

TM stores accepted source+context → target pairs. It is applied automatically:

  • during AI pre-translation — before the model call (free);
  • when opening a new file — you can translate everything from TM in one pass (see below).

The TM key = context (msgctxt) + source + plural. The same text with different contexts keeps different translations.

Translate from TM

Before the AI call the app first walks all strings and applies exact TM matches — the log shows «? From TM free: N strings». This is always profitable: the more accepted translations, the lower the pack spend.

TM suggestions in the editor

The panel under the translation field shows the closest TM matches with a similarity percentage (fuzzy-match). Click to apply. Works offline.

TMX import/export

Menu → Import TM memory / Export TM memory — TMX 1.4b format, compatible with OmegaT, Trados, SmartCat and other CAT tools. Import adds entries (duplicates by key are skipped), export dumps the whole memory. TMX is plain XML — share it with colleagues and between projects.

Storage

TM lives in Data\tm.json (and is mirrored to IndexedDB for fast lookup). The file travels with the program folder — the TM «moves» together with the app.

← Docs