All human-consumable text obtained from AT-SPI calls is expected to be in a localized form. As such, Orca does not do any extra localization processing when working with text obtained via the AT-SPI.
For text generated by Orca itself, Orca handles internationalization and localization using the [gettext>] support of Python. The gettext support of Python is similar to the GNU gettext module. Each human consumable string of Orca is US English text wrapped in a call to gettext.gettext. The call to gettext.gettext will either return a localized string or default to the US English text. Orca depends upon an active and thriving community of open source translators to provide the localizations.
The synthesis of localized speech is to be provided by the underlying gnome-speech implementation.
The generation of localized braille is to be determined. POR HACER: BrlTTY currently does not support this at the moment, but it is expected that the BrlTTY developers will add this in the future.