[← README](../../README.md) · [Docs index](../README.md) · [Customising](README.md) # Languages Both apps, the terminal app and the desktop app, speak 31 languages. They share one set of translations and one setting, so they always agree. By default Coxswain follows your system's language. ![The desktop app in Danish: Hjem, Drev and Git-repositorier in the sidebar, Navn, Størrelse and Ændret as column headers, the F-key bar with Hjælp, Kopiér, Flyt and Afslut, and sizes with a decimal comma](../screenshots/gui-lang-da.png) *Coxswain in Danish. Sizes are written 32,7 KB, with the decimal comma.* ## Contents - [The languages](#the-languages) - [How to use it](#how-to-use-it) - [Which language Automatic picks](#which-language-automatic-picks) - [What you see](#what-you-see) - [Capitals in Greek](#capitals-in-greek) - [Japanese and Korean](#japanese-and-korean) - [Persian, Armenian and Georgian](#persian-armenian-and-georgian) - [Right to left](#right-to-left) - [Settings and config.toml](#settings-and-configtoml) - [In the terminal app](#in-the-terminal-app) - [Questions](#questions) - [Improving a translation](#improving-a-translation) ## The languages Both apps list them by region, and by their own name within a region: | | Language | Code | |---|---|---| | | **Nordic and Baltic** | | |

| Dansk | `da` | |

| Eesti | `et` | |

| Latviešu | `lv` | |

| Lietuvių | `lt` | |

| Norsk (bokmål), Norwegian Bokmål *(new)* | `nb` | |

| Norsk (nynorsk), Norwegian Nynorsk *(new)* | `nn` | |

| Suomi | `fi` | |

| Català | `ca` | |

| Deutsch | `de` | |

| Deutsch (Österreich) | `de-AT` | |

| Deutsch (Schweiz) | `de-CH` | |

| English (United Kingdom) | `en-GB` | |

| Euskara | `eu` | |

| Français | `fr` | |

| Italiano | `it` | |

| Čeština (Czech) | `cs` | |

| Polski (Polish) | `pl` | |

| עברית (Hebrew) | `he` | |

| Հայերեն (Armenian) | `hy` | |

| English (Canada) | `en-CA` | |

| English (Australia) | `en-AU` | |

| English (New Zealand) | `en-NZ` | |

| 日本語 (Japanese) | `ja` | |

| 한국어 (Korean) | `ko` | The ones marked *new* are fresh translations that no native speaker has checked yet: Settings marks them *new* for about a month, four weekly releases after the one that brought them, and [Improving a translation](#improving-a-translation) says how to help. Persian has no state flag here: it is spoken in Iran, Afghanistan and Tajikistan, so it shows a historical banner instead, the Derafsh Shahbaz, the Achaemenid falcon (a gold falcon with spread wings on crimson), drawn for Coxswain. British English is the reference: every text is written in it first. Australian and New Zealand English use British spelling and say "bin"; Canadian English keeps British spelling but says "trash" and uses -ize where British uses -ise. Argentinian Spanish uses *vos*. Austrian and Swiss German are German with their own differences on top; any text they do not change is the German one. Swiss German writes *ss* for every *ß* ("Schliessen", "Grösse") and «guillemets» for „quotes“, as Swiss Standard German does; in the desktop app its numbers are written 1'234.5. Coxswain writes dates as numbers (2026-01-05), so Austrian German has no *Jänner* to show and, for now, no text of its own: it reads as German. ## How to use it **Desktop app:** 1. Open Settings: **Ctrl+,** or **F12**, or **F9** → *Settings*, and choose *Looks*; or start it with `coxswain-gui --settings=language`, which opens *Looks* at *Language*. 2. Under *Language*, click a language. The one in use is shown at the top, with its flag. Below it a field filters the list; the languages are listed under their regions (*Nordic and Baltic*, *Western Europe*, …), each by its own name and flag; a fresh translation has a *new* badge. 3. The window redraws in the new language at once, and the choice is saved to `config.toml`. *Automatic* goes back to following the system; below it, small, is the language that picks now. Under the list, *Translations marked new are fresh…* opens [Improving a translation](#improving-a-translation) in the browser. **Terminal app:** 1. Press **F12** (*12 Settings* on the F-key bar), or start it with `coxswain --settings=language`. 2. In *Looks*, put the cursor on *Language* and press **Enter**: *Automatic* and every language, by its own name. **↑** **↓** and **Enter** take one. 3. Every text changes at once, and the choice is saved to `config.toml`. **By hand:** 1. Open `config.toml` (`coxswain --config-path` prints where). 2. Put `language = "da"` at the top, above the first `[table]`. `"auto"` follows the system. 3. Start the app again (quit with **F10**): a hand edit is read when an app starts. There are no keys for the language. ## Which language Automatic picks With `language = "auto"`, the default, Coxswain reads your system's list of preferred languages (on every system alike, from FreeBSD to Windows), and takes the first one it has, in any regional form (any English counts); otherwise the nearest relative of the first one: | Your system | Coxswain uses | |---|---| | One of the languages above, in any region (`fr-CA`, `sv-FI`, `ca-ES-valencia`, …) | That language | | German in Austria (`de-AT`) | Austrian German | | German in Switzerland or Liechtenstein (`de-CH`, `de-LI`), Swiss German (`gsw`) | Swiss German | | German anywhere else (`de-DE`, `de-LU`, `de-BE`, …) | German | | US English, or English without a region | Canadian English | | Australian, Canadian, New Zealand English | That English | | Any other English (Britain, Ireland, South Africa, India, …) | British English | | Any Spanish (Spain, Mexico, …), Galician and Aragonese | Argentinian Spanish | | Norwegian Bokmål (`nb_NO`, `nb`), and Norwegian with no written standard named (`no`, `no_NO`) | Norsk (bokmål) | | Norwegian Nynorsk (`nn_NO`, `nn`) | Norsk (nynorsk) | | Frisian, Afrikaans | Dutch | | Luxembourgish | German | | Occitan | Catalan | | Polish, Czech, Ukrainian, Greek (`pl_PL`, `cs_CZ`, `uk_UA`, `el_GR`, `el_CY`, …) | That language | | Slovak | Czech, which Slovak readers read | | Japanese, Korean (`ja_JP`, `ko_KR`, …) | That language | | Persian in Iran or Afghanistan (`fa_IR`, `fa_AF`), Dari (`prs`) | Persian | | Armenian, Georgian (`hy_AM`, `ka_GE`) | That language | | Hebrew (also the old code `iw`) | Hebrew | | Anything else | British English | A language code in `config.toml` goes through the same table, so `language = "no"` gives Bokmål and `language = "es"` Argentinian Spanish. ## What you see Everything Coxswain itself says changes: menus, commands and the F-key bar, dialogs, the preview pane and its facts, the duplicate finder, Settings, status messages, errors, sizes (`KB`, or `Ko` in French, with the decimal comma where the language uses one) and the age chips (`5m`, `2d`). Numbers are written the way the language writes them (1 234,5 or 1.234,5). Names in the panels, archives, a comic's pages, Space, the clouds (in the sidebar and under **Alt+F1**) and the chess game table sort by the language's alphabet (Danish *Æ Ø Å* after *Z*, German *Ä* with *A*; [Sorting](../panels/sorting.md)). In Persian, counts, sizes and dates take Persian digits (۳۲٫۷ KB, ۲۰۲۶-۰۱-۰۵) in both apps; paths, keys, versions and `config.toml` values keep their Latin digits. Not translated: - key names (Ctrl, Alt, Enter, F5 …), which are the same in `config.toml` everywhere; - file and folder names, and the names of your favourite groups; - error texts that come from the operating system, which are in the system's own language; - the names of programs and formats (Ollama, LibreOffice, PDF). - *What's new* in Settings → *Overview*, which shows the changelog as it is written, in British English. ## Capitals in Greek Some headings are written in capitals (Settings' area titles, the sidebar's headings, the Cyber theme's dialog titles). Greek drops its accents in capitals and keeps a diaeresis: *Ρυθμίσεις* becomes ΡΥΘΜΙΣΕΙΣ, not ΡΥΘΜΊΣΕΙΣ, and *τσάι* becomes ΤΣΑΪ. The desktop app tells its web view that the page is Greek, which uppercases it so (WebKitGTK on Linux, WebKit on macOS and WebView2 on Windows all do). The terminal app writes only the group headings of Find's results in capitals. Georgian has no capitals in running text, so in Georgian these headings are written as they are, without capitals or wide spacing, in both apps, and the first letter of an error's cause is left as it is. Armenian has capitals and takes them as usual. ## Japanese and Korean Both are written in letters that take two columns in a terminal and need a font that has them. - **Fonts in the desktop app:** after every font list (*Font*, *Monospaced font*, the era fonts of the themes) come Hiragino Sans, Yu Gothic, Meiryo and Noto Sans CJK JP, then Apple SD Gothic Neo, Malgun Gothic and Noto Sans CJK KR (Korean first when Coxswain speaks Korean, as the same Han letter is drawn differently in each). The first installed one draws them: on macOS and Windows one always is. This holds for file names in any language too. - **No font on Linux:** when Coxswain speaks Japanese or Korean and `fc-list` knows no font for it, *Settings → Overview → What's new* says once *No Japanese or Korean font is installed…* and names the package: `noto-fonts-cjk` (Arch), `fonts-noto-cjk` (Debian, Ubuntu), `google-noto-sans-cjk-fonts` (Fedora). Until then the letters show as boxes. - **Typing with an input method** (IME: Mozc, Kotoeri, the Microsoft IME, a Hangul keyboard): works in every text field of the desktop app (Find, the path bar, rename, notes, the dialogs). While a word is being composed, Enter, Esc and the arrows belong to the input method: Enter takes the word and does not run the dialog, Esc drops the word and does not close it. - **Quick search** (**Alt+letter**, then letters) jumps to a name that starts with what you type; a name kept decomposed (as macOS does with Hangul, or か + ゙ for が) is found by the composed letters. An input method composes only in a text field, so in the desktop app quick search takes the letters typed straight from the keyboard; for a Japanese or Korean name use *Find* (**Alt+F7**), where the input method works. In the terminal app, letters an input method commits go to quick search once it is started. - **The terminal app** measures every text by the columns it takes, not by its letters: the column titles, sizes and the marked line are centred and cut to their columns, the F-key bar shares its width out by label, and Find and the dialogs keep their frames. Your terminal needs a font with these letters (most terminals fall back to one by themselves). ## Persian, Armenian and Georgian ![The desktop app in Persian (Cyber), mirrored: the sidebar on the right with خانه and سیستم, the panes' columns running right to left, sizes and ages in Persian digits, the F-key bar from F1 راهنما on the right to F10 خروج on the left](../screenshots/gui-lang-fa.png) - **Fonts in the desktop app:** after every font list come Vazirmatn, Noto Sans Arabic and Noto Naskh Arabic (Geeza Pro on macOS; Segoe UI and Tahoma on Windows have the letters), Noto Sans Armenian and Noto Sans Georgian (Sylfaen on Windows; macOS has its own), then the Japanese and Korean ones. On FreeBSD they come with `noto` (`pkg install noto`), on NetBSD with `noto-ttf` (`pkgin install noto-ttf`), on OpenBSD with `noto-fonts` (`pkg_add noto-fonts`), and on Linux with `noto-fonts` (Arch), `fonts-noto-core` (Debian, Ubuntu) or `google-noto-sans-arabic-fonts`, `google-noto-sans-armenian-fonts` and `google-noto-sans-georgian-fonts` (Fedora); Vazirmatn, drawn for Persian, is `fonts-vazirmatn` on Debian and Ubuntu. When Coxswain speaks one of these and `fc-list` knows no font for it, *Settings → Overview → What's new* says once *No font for فارسی is installed…* with these packages. - **Persian in the desktop app** is right to left, laid out as Hebrew is (see [Right to left](#right-to-left)); its letters join as they should, and headings that other languages write in spaced capitals are written plain, as spacing pulls joined letters apart. - **Persian in the terminal app:** a terminal draws a letter per cell; whether Persian letters join and run right to left is up to the terminal, not Coxswain. Konsole and mlterm join them and run them right to left; Windows Terminal joins them; GNOME Terminal and other VTE terminals do both with bidi on. Alacritty, kitty, foot, WezTerm, xterm and the macOS Terminal show them unjoined and in reversed order. Coxswain does no shaping of its own: the terminal gets the letters in their logical order. - **Armenian and Georgian** are left to right and need nothing but a font; in the terminal app most terminals fall back to one by themselves. ## Right to left ![The desktop app in Hebrew, mirrored: the sidebar on the right, file names aligned right with their icons on the right, the F-key bar running from F1 on the right](../screenshots/gui-lang-he.png) *Coxswain in Hebrew: the whole layout mirrors.* In Hebrew and Persian the desktop app mirrors its layout: the sidebar is on the right, text is aligned to the right, the columns run from right to left, and the back arrow points right. File names, paths and commands stay left to right inside it, as Hebrew and Persian systems show them, and notes and the command line take the direction of what you type. The terminal app is limited by the terminal: most terminals, Alacritty among them, do not do right-to-left text, so Hebrew and Persian there show their letters in reversed order. Terminals that support it (for example Konsole, mlterm, Windows Terminal, or GNOME Terminal with bidi on) show it correctly; for Persian see also [Persian, Armenian and Georgian](#persian-armenian-and-georgian). ## Settings and config.toml | Setting | Key | Type, default | |---|---|---| | Settings → *Looks* → *Language* | `language` | text: `"auto"` or a code from the table; `"auto"` | One key for both apps. ## In the terminal app The same 31 languages and the same translations, chosen by the same `language` key. Its own Settings pick it at once (**F12** → *Looks* → *Language*); a change made in the desktop app or in the file is read when it starts. `coxswain --languages` prints the list by region, the one in use and the new ones marked, and where to suggest a better word. Right to left depends on the terminal (see [Right to left](#right-to-left)). The help (**F1**), the command list (**F9**), dialogs, the F-key bar and status texts are all translated. ## Questions #### My system is in US English. Why does Coxswain say "colour" but "trash"? US English, and English without a region, get Canadian English: British spelling, North American words. Pick *English (United Kingdom)* in Settings for "bin", or keep it. #### I changed the language in Settings and the terminal app still speaks the old one. You changed it in the desktop app's Settings: a terminal app already running reads `language` when it starts. Quit it (**F10**) and start it again, or pick the language in its own Settings (**F12** → *Looks* → *Language*), which changes it at once. #### Why did my files change their order when I changed the language? Names sort by the alphabet of the language Coxswain speaks, so Danish puts *Æble* after *Zebra* and British English puts it with *A*. A folder read after the change (open it again, or **Ctrl+R**) shows the new order ([Sorting](../panels/sorting.md#why-do-æ-ø-and-å-come-after-z-and-ä-with-a)). #### How do I go back to following the system? Settings → *Looks* → *Language* → *Automatic*, or `language = "auto"` in `config.toml` (or delete the line). #### My system lists British English first and Danish second, and Coxswain speaks Danish. That is how *Automatic* works now: British English is also what an unknown language falls back to, so it passes over every system language that comes out as British English (`en-GB`, `en-IE`, `en-IN`, …) and takes the first that gives another language; only when there is none does it use British English. Set `language = "en-GB"` (or pick *English (United Kingdom)* in Settings) to keep English. #### My system is set to German (Switzerland). Why does Coxswain write "Grösse"? Swiss Standard German has no *ß*: a system locale `de_CH` (or `de_LI`, or Swiss German `gsw`) picks *Deutsch (Schweiz)*, which writes *ss* and «guillemets». For *ß* and „quotes“ pick *Deutsch* in Settings → *Looks* → *Language*, or set `language = "de"`. #### Polish, Czech, Ukrainian, Greek, Persian, Armenian, Georgian, Japanese or Korean reads oddly in places. Why? These translations have not been read by a native speaker yet. While a translation is new (about a month, four weekly releases), Settings marks it *new* and Coxswain shows the notice "*Polski is a new translation…*" once (under Settings → *Overview* → *What's new* in both apps; the terminal app also says it in the status line). The mark goes when the month is over, not because someone checked it. A better word is very welcome: see [Improving a translation](#improving-a-translation). #### My system is Norwegian. Do I get Bokmål or Nynorsk? The one your system names: `nb_NO` (or `nb`) gives *Norsk (bokmål)*, `nn_NO` (or `nn`) gives *Norsk (nynorsk)*. A system that says only `no` or `no_NO` gets Bokmål, the one most Norwegians write. To pick the other, choose it in Settings → *Looks* → *Language*, or set `language = "nn"` (or `"nb"`) in `config.toml`. Both are new and marked *new* in Settings until 15 November 2026; a native reader's corrections are very welcome (see [Improving a translation](#improving-a-translation)). #### Why does Persian have a falcon and not a flag? Persian is the language of Iran, Afghanistan and Tajikistan; no one state flag stands for it. Coxswain shows the Derafsh Shahbaz, the falcon banner of the Achaemenids, drawn for Coxswain (MIT licence, `docs/flags/derafsh.svg`). #### Persian letters in the terminal app are separate and backwards. Can Coxswain fix that? No: joining letters and running them right to left is the terminal's work. Use a terminal that does it: Konsole or mlterm, or GNOME Terminal with bidi on (Windows Terminal joins them). Alacritty, kitty, foot, WezTerm and xterm do not. The desktop app shows Persian correctly everywhere. #### Why are sizes in Persian written ۳۲٫۷ but the path keeps 2026? Counts, sizes and dates are numbers to read, so they take Persian digits. A path, a key, a version or a `config.toml` value must stay as typed and as the system has it, so its digits stay Latin. #### Why are the Cyber theme's headings not in capitals in Georgian? Georgian has no capitals in running text. Unicode's capitals for Mkhedruli (Mtavruli) are for rare all-caps lettering and most fonts lack them, so they would show as boxes. Coxswain writes Georgian headings as they are. #### Persian, Armenian or Georgian shows as boxes. What is missing? A font with those letters. On FreeBSD `pkg install noto`, on NetBSD `pkgin install noto-ttf`, on OpenBSD `pkg_add noto-fonts`; on Linux install `noto-fonts` (Arch), `fonts-noto-core` (Debian, Ubuntu) or `google-noto-sans-arabic-fonts`, `google-noto-sans-armenian-fonts`, `google-noto-sans-georgian-fonts` (Fedora). Then start the desktop app again. The desktop app says so once under *Settings → Overview → What's new*. See [Persian, Armenian and Georgian](#persian-armenian-and-georgian). #### Japanese or Korean shows as boxes. What is missing? A font with those letters. On Linux install `noto-fonts-cjk` (Arch), `fonts-noto-cjk` (Debian, Ubuntu) or `google-noto-sans-cjk-fonts` (Fedora) and start the desktop app again; the desktop app says so once under *Settings → Overview → What's new*. In the terminal app, it is the terminal's font that needs them. See [Japanese and Korean](#japanese-and-korean). #### I type Japanese and press Enter, and the dialog does nothing. Why? The first Enter takes the word your input method was composing; it is not a key for Coxswain yet. Press Enter once more to run the dialog. Esc likewise drops the composed word first. #### Why are key names in English? Keys are written the same everywhere (Ctrl, Alt, F5), so a key in `config.toml` means the same in every language, and the names match what is printed on most keyboards. #### Can I type Alt Gr letters (@, {, €, ą) in Find, rename and the command line? Yes, in both apps, on any keyboard layout. A Windows terminal sends an Alt Gr letter as **Ctrl+Alt** with the letter; the terminal app types it, and keeps **Ctrl+Alt** with a plain letter or digit for shortcuts. #### Does the language change what Find reads in pictures? Yes: [text recognition](../search/scans.md) reads English and your system's language, when tesseract has that language installed. [Search by meaning](../search/meaning.md) works across languages whatever you pick: a query in English finds a Danish document. #### The F-key bar in my language is cut short. Is that a mistake? In the terminal app at 80 columns, eleven F-keys share the bar: about six columns a label (nine at 120). When a label does not fit its slot, short labels lend their spare room to long ones, so only the longest are cut, marked with *…*; a wider window shows them whole. The terminal app keeps Norton Commander's "RenMov", "Mkdir" and "PullDn" in English on purpose (Persian, Armenian and Georgian have short words of their own); the desktop app says "Move", "New folder" and "Commands". A better short word is welcome; see [Improving a translation](#improving-a-translation). ## Improving a translation The translations were written with care but have not all been checked by native speakers yet. The ones marked *new* (Polish, Czech, Ukrainian, Greek, Japanese, Korean, Persian, Armenian, Georgian, Norwegian Bokmål and Nynorsk) need a native reader most, then Basque, Latvian and Lithuanian. Corrections are very welcome, in either of two ways: - **Tell us:** open an [issue on GitHub](https://github.com/mwo-dk/coxswain/issues/new) with the language, the text as it is (or where it shows) and what it should say. No setup needed. - **Change it yourself,** in a pull request: 1. Each language is one file in [`crates/coxswain-core/locales/`](../../crates/coxswain-core/locales/), for example `da.json`. It maps a key to its text: ```json "dupes.keep_newest": "Markér alle undtagen den nyeste", "items": { "one": "{n} element", "other": "{n} elementer" } ``` 2. Change the text, keeping every `{placeholder}` (you may move it). Texts with a count have one entry per plural form your language uses: `one`/`other` for most, `one`/`few`/`other` for Lithuanian, `zero`/`one`/`other` for Latvian, `one`/`two`/`other` for Hebrew, and `one`/`many`/`other` for French, Italian, Spanish and Catalan, `one`/`few`/`many`/`other` for Polish and Ukrainian, `one`/`few`/`other` for Czech, `one`/`other` for Persian and Armenian (where 0 is `one` too) and for Georgian, and only `other` for Japanese and Korean, which do not change a word for a count. 3. Run `cargo test -p coxswain-core i18n`: it checks that every file parses, has no unknown keys, keeps the placeholders and has an `other` form. 4. Open a pull request. **Adding a language:** add its code, name, flag and region to `LANGUAGES` and `source()` in [`crates/coxswain-core/src/i18n.rs`](../../crates/coxswain-core/src/i18n.rs), map its system codes in `nearest()`, give it plural rules in `plural()` if it needs other than one/other, copy `en-GB.json` to the new file and translate it, and add the flag (from [flag-icons](https://github.com/lipis/flag-icons)) to `docs/flags/` and to the list in `gui/src/Settings.svelte`. Flags: [flag-icons](https://github.com/lipis/flag-icons), MIT licence (`docs/flags/LICENSE`); Persian's falcon banner (`derafsh.svg`) is drawn for Coxswain, MIT licence. --- [← Previous: Your own theme and the colour slots](own-theme.md) · [Next: Changing keys →](keys.md)