So it seems that there's technically no difference as of Django 1.5. Template engine internally marks a variable for translation (by setting its translate attribute) in two cases:

  • when you do {% trans VAR %} (see TranslateNode), or
  • if the name of a variable starts with _( and ends with ) (see Variable.__init__).

Later, when the variable is being resolved, Django wraps it with ugettext or pgettext if it sees the translate attribute.

However, as can be seen from source code, there are some flexibility considerations in favor of {% trans %} tag:

  • you can do {% trans "String" noop %}, which will put the string for translation into .po files, but won't actually translate the output when rendering (no internal translate attribute on variable, no ugettext call);
  • you can specify message context*, like {% trans "May" context "verb" %};
  • you can put translated message into a variable for later use*, like {% trans "String" as translated_string %}.

* As of Django 1.4.

Please feel free to correct me or post a better answer in case I'm missing anything.


The trans template tag calls the ugettext() function. In Django _() is an alias to ugettext(). This is covered in the django docs.