no fixed listany BCP-47 tag ICU knows

Ship the whole product, not the English half of it.

every swap re-runs the gate

placeholders

plurals

glossary

length

Transglot translates into any language you can name with a BCP-47 tag. There is no supported-languages list to check against, because there is no list: type the tag and the name, the plural rules and the writing direction are worked out on the spot.

icu locales accepted708
cldr plural languages224
plural forms in arabic6

the atlas

Type one sentence. See what each language needs from it.

pick up to six tiles below. Each one opens on the right: writing direction, the plural forms that language requires, the placeholders that have to survive, and the structure a translation must keep. Nothing is translated and nothing leaves the page

atlas · languages consoleacme / inbox / inbox.count0 / 6

sentence

0 of 142 selected · 0 plural forms required in total · 1 placeholder

the atlasone tile per language · grouped by script · the small figure is how many plural forms that language has

Latin89

Cyrillic15

Arabic script7

Han, Kana and Hangul3

Indic scripts14

other scripts14

solid: this language has published plural rulesdashed: no published plural rule, so every number takes one form

what this language needsplaceholders · plurals · glossary · length

nothing selected_
pick a tile on the atlas. Each language opens here: writing direction, plural forms required, placeholders to carry, and the structure the translation must keep

This is a reading of your sentence, not a translation: nothing here is sent anywhere and no draft is produced. Plural forms and writing directions come from the sampled locales further down this page, matched to a tile by base language, and for the rest from the CLDR plural rules ICU ships, written out here by hand and checkable against CLDR. A language CLDR has no rule for is drawn dashed: ICU uses other alone. Tiles outside the sample are grouped by the script their name is printed in. Placeholders, plural arguments and the structure are read out of the sentence as you type.

counted, not claimed

The four numbers we can actually count.

none of them is a total, because a total would be wrong in both directions: too small for what the validator accepts, and too precise for a thing with no boundary

icu locales

708

the table a tag is checked against, before a structural fallback widens it further. Symfony\Component\Intl\Locales::getLocales().

cldr plural languages

224

languages with their own plural rules. A translation missing a form its language needs fails the gate instead of shipping. Straight from the CLDR plural tables.

sampled below

rows in the table further down this page. A sample of where teams start, not a catalogue of what is allowed.

plural forms, arabic

zero, one, two, few, many and other. Arabic needs all six, which is the case a hand-written i18n file usually gets wrong.

one tile per language

The {count} base languages ICU resolves on its own.

regional variants are left off so the wall reads as one tile per language rather than one language many ways · the dashed tile is the rest of the ICU table

+566

the sample

What you get the moment you add a locale.

the native name, the writing direction, the plural forms and the script, worked out for you rather than typed in · filter by region

de
Deutsch
German, written in Latin.
2 plural forms
one · other
scored regression case
fr
français
French, written in Latin.
3 plural forms
one · many · other
scored regression case
fr-CA
français (Canada)
French (Canada), written in Latin.
3 plural forms
one · many · other
es
español
Spanish, written in Latin.
3 plural forms
one · many · other
scored regression case
es-419
español (Latinoamérica)
Spanish (Latin America), written in Latin.
3 plural forms
one · many · other
pt-PT
português (Portugal)
Portuguese (Portugal), written in Latin.
3 plural forms
one · many · other
pt-BR
português (Brasil)
Portuguese (Brazil), written in Latin.
3 plural forms
one · many · other
it
italiano
Italian, written in Latin.
3 plural forms
one · many · other
nl
Nederlands
Dutch, written in Latin.
2 plural forms
one · other
sv
svenska
Swedish, written in Latin.
2 plural forms
one · other
da
dansk
Danish, written in Latin.
2 plural forms
one · other
nb
norsk bokmål
Norwegian Bokmål, written in Latin.
2 plural forms
one · other
fi
suomi
Finnish, written in Latin.
2 plural forms
one · other
is
íslenska
Icelandic, written in Latin.
2 plural forms
one · other
pl
polski
Polish, written in Latin.
4 plural forms
one · few · many · other
cs
čeština
Czech, written in Latin.
4 plural forms
one · few · many · other
sk
slovenčina
Slovak, written in Latin.
4 plural forms
one · few · many · other
hu
magyar
Hungarian, written in Latin.
2 plural forms
one · other
ro
română
Romanian, written in Latin.
3 plural forms
one · few · other
hr
hrvatski
Croatian, written in Latin.
3 plural forms
one · few · other
sl
slovenščina
Slovenian, written in Latin.
4 plural forms
one · two · few · other
sr
српски
Serbian, written in Cyrillic.
3 plural forms
one · few · other
bg
български
Bulgarian, written in Cyrillic.
2 plural forms
one · other
ru
русский
Russian, written in Cyrillic.
4 plural forms
one · few · many · other
scored regression case
uk
українська
Ukrainian, written in Cyrillic.
4 plural forms
one · few · many · other
el
Ελληνικά
Greek, written in Greek.
2 plural forms
one · other
tr
Türkçe
Turkish, written in Latin.
2 plural forms
one · other
lt
lietuvių
Lithuanian, written in Latin.
4 plural forms
one · few · many · other
lv
latviešu
Latvian, written in Latin.
3 plural forms
zero · one · other
et
eesti
Estonian, written in Latin.
2 plural forms
one · other
ca
català
Catalan, written in Latin.
3 plural forms
one · many · other
ar
العربية
Arabic, written in Arabic, right to left.
6 plural forms
zero · one · two · few · many · other
scored regression case
he
עברית
Hebrew, written in Hebrew, right to left.
3 plural forms
one · two · other
fa
فارسی
Persian, written in Arabic, right to left.
2 plural forms
one · other
sw
Kiswahili
Swahili, written in Latin.
2 plural forms
one · other
af
Afrikaans
Afrikaans, written in Latin.
2 plural forms
one · other
am
አማርኛ
Amharic, written in Ethiopic.
2 plural forms
one · other
hi
हिन्दी
Hindi, written in Devanagari.
2 plural forms
one · other
bn
বাংলা
Bangla, written in Bangla.
2 plural forms
one · other
ta
தமிழ்
Tamil, written in Tamil.
2 plural forms
one · other
te
తెలుగు
Telugu, written in Telugu.
2 plural forms
one · other
mr
मराठी
Marathi, written in Devanagari.
2 plural forms
one · other
ur
اردو
Urdu, written in Arabic, right to left.
2 plural forms
one · other
ja
日本語
Japanese, written in Han + Kana.
1 plural forms
other
scored regression case
ko
한국어
Korean, written in Hangul.
1 plural forms
other
zh-Hans
中文（简体）
Chinese (Simplified), written in Han (Simplified).
1 plural forms
other
zh-Hant
中文（繁體）
Chinese (Traditional), written in Han (Traditional).
1 plural forms
other
th
ไทย
Thai, written in Thai.
1 plural forms
other
vi
Tiếng Việt
Vietnamese, written in Latin.
1 plural forms
other
id
Indonesia
Indonesian, written in Latin.
1 plural forms
other
ms
Melayu
Malay, written in Latin.
1 plural forms
other
fil
Filipino
Filipino, written in Latin.
2 plural forms
one · other
ka
ქართული
Georgian, written in Georgian.
2 plural forms
one · other
hy
հայերեն
Armenian, written in Armenian.
2 plural forms
one · other

showing 54 of 54 sampled locales · the product accepts many more than either number

a locale that is not in this sample behaves the same way: the same four values, worked out the same way, the moment you add the tag.

FAQs

What people ask before adding the first locale.

Which languages does Transglot support?

Any language you can name with a BCP-47 tag. There is no allowlist to be on: a tag is checked against the ICU locale table, which holds 708 entries, and a structural fallback accepts real tags beyond even that. If ICU knows your tag, the product takes it.

How many languages is that, exactly?

There is no honest single number. Any figure would be too small for what the validator accepts and too precise for a thing with no boundary. The countable numbers are 708 locales in the ICU table and 224 languages with their own plural rules.

Does it handle right-to-left languages like Arabic and Hebrew?

Yes. Writing direction is read from ICU per locale rather than kept in a list someone has to maintain, so Arabic, Hebrew, Persian, Urdu, Dhivehi and N’Ko all come out right to left without anyone adding them anywhere. There are eleven right-to-left scripts, and the editor renders the target in its own direction while you work on it.

How does it handle plurals in languages with more than two forms?

It loads the CLDR plural table for the target language, from the same data set the quality gate checks against. Arabic gets all six forms, Polish gets its three, Japanese gets one. A translation missing a form its language requires fails the gate instead of shipping a screen that reads wrong once every six numbers.

Can I use regional variants like pt-BR, es-419 or en-GB?

Yes. A regional variant is its own locale, so pt-BR and pt-PT can carry different wording and different glossary terms. Tags are canonicalised on the way in, so pt_br, PT-br and pt-BR all land on the same locale rather than creating three.

Does it support both Simplified and Traditional Chinese?

Yes, as two locales: zh-Hans and zh-Hant. Script is part of the tag, so they are translated separately and reviewed separately rather than converted from one another.

How many languages can I have on the free plan?

Fifteen locales in one project, and that count includes your source language, so 14 targets. The free plan also carries 100,000 words to start, 10,000 a month after, and three seats. Paid plans do not cap locales.

How do I add a new language?

Add the tag to the project. Nothing else has to be configured: the display name, the endonym, the plural forms and the writing direction all come from the tag, and the next run drafts every string that has no translation in it yet.

What about languages that do not use spaces, like Japanese or Thai?

They work the same way. Counting is script aware, so a Japanese or Chinese string is measured by its script rather than by counting spaces, which is what stops CJK from being billed as one enormous word.

Is quality the same in every language?

Every locale goes through the same engine, the same prompt and the same four gates. We publish no per-language quality grade, because we do not measure one. Some of the sampled locales carry a scored regression case in our evaluation corpus, and the table above marks which; that means a regression there is caught automatically, not that the language is better.

What happens if a tag is wrong?

It is refused at the point you add it, not silently accepted. Made-up tags like en-XX do not pass, and neither does a real language paired with a region or script that does not exist.

Can I translate from a source language other than English?

Yes. The source is whichever locale the project marks as source, and every other locale in the project is drafted from it. Nothing assumes English.

works with what you already run