OpenMW supports localisation of mods using ICU MessageFormat wrapped in YAML. Currently this is only possible using the openmw.core.l10n lua function.


Locales usually have the form {lang}_{COUNTRY}, where {lang} is a lowercase two-letter language code and {COUNTRY} is an uppercase two-letter country code. Localisation files must have this exact capitalisation and separator to be recognized.

When users request a locale using the preferred locales setting they do not need to match capitalisation and can also use hyphens instead of underscores. The locale will be normalised to the above format.

Locales may also contain variants and keywords, though these usually will not be necessary. See The Locale chapter of the ICU Guide for full details.


When OpenMW looks up messages at runtime, it starts with the first requested locale, and then looks at that locale’s more generic ancestors before looking at the next requested locale. E.g. en_GB_OED will fall back to en_GB, which will fall back to en.

When including localisations with specific country variants (or more specific variants/keywords), you should always include the more generic version as well.

E.g. if you include en_US.yaml and en_GB.yaml localisation files, you should also include en.yaml, since other English locales will fall back to that (e.g. en_CA, en_AU, en_NZ). You can put an arbitrary en locale of your choice in en.yaml, and then leave the file for that variant empty (since all lookups for the variant will fall back to en anyway).

Note that because of the fallbacks only messages which differ between variants need to be included in the country-specific localisation files.

Localisation Files

Localisation files (containing the message names and translations) should be stored in the VFS as files of the form l10n/<ContextName>/<Locale>.yaml.

Naming policy

“ContextName” should be in line with Lua scripts naming policy:

  • L10n files for scripts/<ModName>/<ScriptName>.lua should be placed to l10n/<ModName>/<Locale>.yaml.
  • L10n files for scripts/<AuthorName>/<ModName>/<ScriptName>.lua should be placed to l10n/<AuthorName><ModName>/<Locale>.yaml.

In most cases one mod should have only one l10n context. Don’t create a new context for each single message. Really big mods with hundreds and thousands of messages can have several l10n contexts. In this case all context names should start with the name of the mod. I.e. <ContextName> = <ModName><Subcontext> (or <AuthorName><ModName><Subcontext>).

L10n contexts with prefix “OMW” are reserved for the OpenMW itself (in particular for built-in scripts scripts/omw/) and shouldn’t be used in mods.

Built-in l10n contexts “Interface” and “Calendar” don’t have the “OMW” prefix because these messages are more generic and can be reused in mods.


Messages contents have the form of ICU MessageFormat strings. See the Formatting Messages chapter of the ICU Guide for a guide to MessageFormat, and see The ICU APIdoc for full details of the MessageFormat syntax.


good_morning: 'Good morning.'

you_have_arrows: |-
  {count, plural,
    one {You have one arrow.}
    other {You have {count} arrows.}
good_morning: "Guten Morgen."
you_have_arrows: |-
  {count, plural,
    one {Du hast ein Pfeil.}
    other {Du hast {count} Pfeile.}
"Hello {name}!": "Hallo {name}!"

Select rules can be used to match arbitrary string arguments. The default keyword other must always be provided.

pc_must_come: {PCGender, select,
    male {He is}
    female {She is}
    other {They are}
  } coming with us.

Numbers have various formatting options and can also be formatted with custom patterns. See The ICU Guide

quest_completion: "The quest is {done, number, percent} complete."
# E.g. "You came in 4th place"
ordinal: "You came in {num, ordinal} place."
# E.g. "There is one thing", "There are one hundred things"
spellout: "There {num, plural, one{is {num, spellout} thing} other{are {num, spellout} things}}."
numbers: "{int} and {double, number, integer} are integers, but {double} is a double"
rounding: "{value, number, :: .00}"