Merge existing translations from an old translation file to a new one as well as fill any missing translations from translation memory via fuzzy matching.

This functionality used to be part of pot2po and corresponds to “msgmerge” from the gettext package.

pretranslate works on PO and XLIFF files.


pretranslate [options] <input> <output>


<input> is the translation file or directory to be pretranslated
<output> is the translation file or a directory where the pretranslated version will be stored


--version show program’s version number and exit
-h, --help show this help message and exit
--manpage output a manpage based on the help
 show progress as: dots, none, bar, names, verbose
 show errorlevel as: none, message, exception, traceback
-i INPUT, --input=INPUT
 read from INPUT in pot format
-x EXCLUDE, --exclude=EXCLUDE
 exclude names matching EXCLUDE from input paths
-o OUTPUT, --output=OUTPUT
 write to OUTPUT in po, pot formats
-t TEMPLATE, --template=TEMPLATE
 read old translations from TEMPLATE
-S, --timestamp
 skip conversion if the output file has newer timestamp
--tm=TM The file to use as translation memory when fuzzy matching
 The minimum similarity for inclusion (default: 75%)
 Disable all fuzzy matching


pretranslate -t zu-1.0.1 -tm zu_tm.po zu-2.0.2 zu-2.0.2-translated

Here we are pretranslating the PO or XLIFF files in zu-2.0.2 using the old translations in zu-1.0.1 and fuzzy matches from the zu_tm.po compendium. the result is stored in zu-2.0.2-translate

Unlike pot2po pretranslate will not change anything in the input file except merge translations, no reordering or changes to headers.


It helps to understand when and how pretranslate will merge. The default is to follow msgmerge’s behaviour but we add some extra features with fuzzy matching:

  • If everything matches we carry that across
  • We can resurrect obsolete messages for reuse
  • If we cannot find a match we will first look through the current and obsolete messages and then through any global translation memory
  • Fuzzy matching makes use of the Levenshtein distance algorithm to detect the best matches


Fuzzy matches are usually of good quality. Installation of the python-Levenshtein package will speed up fuzzy matching. Without this a Python based matcher is used which is considerably slower.