Reference data
Each language page has the sources, the frequency bands, and the downloads for that pack. The methodology page explains the method. The free resources page has files that you can download now.
Mandarin → English
中文
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Mandarin. Chinese writes no spaces, so the reader splits each sentence into words before you tap one: 我喜欢读书 reads as 我 · 喜欢 · 读书, not five loose characters. Because a character does not show how it sounds, pinyin sits above every word while you read, and each reading retires once you mark that word known. Entries are keyed on Simplified, and Traditional resolves to the same entry, so 這 and 这 both answer zhè.
View reference data🇮🇳Hindi → English
हिन्दी
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Hindi. Devanagari writes a space between words, so a tap on किताब reaches the entry with no extra engine, and combining marks stay inside the word. NFC folds the two nukta spellings (क़ and क + ◌़) onto one key, and the danda । ends a sentence. One honest caveat: the sentence bank is smaller than the European packs, because Tatoeba holds fewer Hindi–English pairs. It covers about 72% of the two thousand most useful practice words. This is the Devanagari register. Urdu writes the same spoken language in a different script and needs its own pack.
View reference data🇪🇸Spanish → English
Español
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Spanish. We don't ship our own Spanish Anki deck; you mine your own cards as you read, and the community deck below covers the vocabulary foundation.
View reference data🌍Arabic → English
العربية
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Arabic: the first right-to-left language pack. Modern Standard Arabic, the written form newspapers and Wikipedia use, not a spoken dialect. A dictionary prints كَتَبَ and a newspaper prints كتب, so entries are keyed on the unvocalized spelling and running text resolves without the vowel marks, while the vocalized spelling shows in the definition. The four alef spellings fold to one key, so إلى and الى answer the same entry, and lookups peel the articles, conjunctions and pronoun endings that Arabic writes solid, so a tap on وبالقلم reaches the entry for قلم on its own. Typed practice accepts an unvocalized answer, because that is how the language is written.
View reference data🇫🇷French → English
Français
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in French. We don't ship our own French Anki deck; you mine your own cards as you read, and the community deck below covers the vocabulary foundation.
View reference data🌍Bengali → English
বাংলা
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Bengali: the first Brahmic-script pack. Bengali attaches its case endings, plurals and classifiers to the end of a word with no space, so lookups peel them to reach the entry (সালের → সাল, বইগুলোর → বই), and the danda । ends a sentence. One honest caveat: the dictionary is small, because English Wiktionary holds far fewer Bengali entries than it does for the European packs. It answers about 88% of the thousand most common words and thins out over formal and news vocabulary, where a tap falls through to AI translation instead. Spoken with an Indian Bengali voice, because that is the only Bengali voice Google publishes.
View reference data🇧🇷Portuguese → English
Português
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Portuguese. Brazilian Portuguese (pt-BR) by default, with a Tatoeba sentence bank and an on-device dictionary covering rich Portuguese inflection (nasal vowels, conjugations, the personal infinitive).
View reference data🇷🇺Russian → English
Русский
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Russian: the first Cyrillic language pack. The on-device dictionary covers Russian's rich inflection (six cases, verb aspect pairs, suppletive forms like шёл → идти), understands the everyday е-for-ё spelling (теплое finds тёплый), and shows lexical stress in definitions.
View reference data🇮🇩Indonesian → English
Bahasa Indonesia
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Indonesian. Official spelling has no diacritics and almost no inflection. Reduplicated plurals stay one word (buku-buku), and lookups peel voice prefixes and possessive clitics (membeli → beli, namanya → nama).
View reference data🇩🇪German → English
Deutsch
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in German. We don't ship our own German Anki deck; the excellent free community decks below cover flashcards.
View reference data🇹🇷Turkish → English
Türkçe
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Turkish. Agglutination is the whole problem here: a single word stacks plural, possessive and case endings (evlerimizden, “from our houses”), and the on-device dictionary unstacks them back to the lemma. Lookups also handle the dotted and dotless i correctly, so a sentence-initial İyi finds iyi and ILIK (“lukewarm”) never collapses onto İLİK (“marrow”).
View reference data🇮🇹Italian → English
Italiano
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Italian. The on-device dictionary covers rich Italian inflection, while the Tatoeba sentence bank keeps elisions such as l'acqua and un'amica intact in practice.
View reference data🇵🇱Polish → English
Polski
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Polish. The on-device dictionary covers seven cases across three genders, verb aspect pairs, and the consonant alternations that defeat naive stemming: noga becomes nodze, miasto becomes mieście, pies becomes psowi. Definitions also carry IPA, so tapping a word shows how to say it as well as what it means.
View reference data🇺🇦Ukrainian → English
Українська
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Ukrainian, as its own language rather than an approximation of Russian. Ukrainian spelling treats the apostrophe as a letter (п'ять, м'ясо, з'їзд), so the reader keeps those words whole and tappable instead of splitting them in half, and a lookup finds the entry whichever apostrophe your source typed. The alphabet includes ґ, є, і and ї, and definitions are Ukrainian ones — так is “yes” here, not the Russian “like that”.
View reference data🇳🇱Dutch → English
Nederlands
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Dutch. We don't ship our own Dutch Anki deck; you mine your own cards as you read, and the community deck below covers the vocabulary foundation.
View reference data🇿🇦Afrikaans → English
Afrikaans
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Afrikaans. The only pack with its own downloadable Anki deck and published frequency list, built from a blend of Afrikaans Wikipedia (written) and OpenSubtitles (spoken).
View reference data🇨🇿Czech → English
Čeština
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Czech. Vowel length carries meaning here, so byt (“a flat”) and být (“to be”) stay two separate words instead of collapsing onto one key. The on-device dictionary covers seven cases, verb aspect pairs, and the stem changes that defeat naive stemming: kůň becomes koňmi, pes becomes psovi, město becomes městě. Definitions also carry IPA, so tapping a word shows how to say it as well as what it means.
View reference data🇸🇪Swedish → English
Svenska
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Swedish. Nouns take a definite suffix (hus → huset → husen), and two genders split the indefinite article (en bok, ett hus). The on-device dictionary carries those forms plus the å/ä/ö letters. Solid compounds stay one word (sjukhusparkering).
View reference data🇬🇷Modern Greek → English
Ελληνικά
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Modern Greek. This pack is monotonic Greek. The Koine Greek pack stays separate for the New Testament. A lookup folds the tonos. Typed practice accepts an answer with no marks. One honest caveat: Tatoeba holds fewer Greek–English pairs than the European packs. The bank covers about 81% of the two thousand most useful practice words.
View reference data🇭🇺Hungarian → English
Magyar
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Hungarian. Hungarian adds case and possession to the noun. A lookup maps those forms to the lemma. Verbs mark a definite object. The on-device dictionary carries those forms plus the ő and ű letters.
View reference data🇫🇮Finnish → English
Suomi
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Finnish. Finnish puts meaning in the case ending. A lookup maps kirjan back to kirja. The on-device dictionary covers the fifteen cases plus the ä and ö letters.
View reference data🟩Esperanto → English
Esperanto
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Esperanto. Because the orthography is fully phonemic, every lookup carries a rule-generated IPA pronunciation, and audio is spoken by eSpeak NG — the only text-to-speech engine that exists for Esperanto — hosted by Lector on every plan. Lookups understand Esperanto's regular word-building (malsanulejo → mal- + san + -ul- + -ej-) and accept x-system typing (gxardeno → ĝardeno).
View reference data🏛️Koine Greek → English
Κοινή
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in polytonic Koine. Built for the Greek New Testament: frequency ranks come from the GNT itself (the ~300 most frequent lemmas cover most of the running text), cloze sentences are verse-aligned with their references kept, lookups forgive accent variance (τὸν finds τόν) and typed practice accepts unaccented answers. No synthesized audio on principle: Koine pronunciation is reconstructed and disputed, so the speaker stays silent rather than guessing.
View reference data🏛️Latin → English
Latina
Fully supported — read, tap-to-define, mine vocabulary, and practice frequency-banded cloze in Latin. Dictionaries mark vowel length; running text almost never does, so amāre and amare are one key, and typed practice accepts the unmarked form. Lookups also try the u/v and i/j spellings editions mix (uult finds vult). Frequency ranks come from the Dickinson College Commentaries Latin Core Vocabulary. No synthesized audio on principle: classical and ecclesiastical pronunciation are disputed, so the speaker stays silent rather than guessing.
View reference data🇯🇵Japanese → English
日本語
Reader and on-device dictionary ship today. Furigana sits above kanji, and a morphological analyser keeps conjugated verbs whole. The cloze bank is the remaining work.
View reference data🇰🇷Korean → English
한국어
Reader and on-device dictionary ship today. Lookups peel postpositions and connective endings off an eojeol (도서관에서 → 도서관). The cloze bank is the remaining work.
View reference dataVietnamese → English
Tiếng Việt
On the roadmap. The reader needs word-boundary support first, because a Vietnamese word is often two syllables with a space between them: học sinh is one word, not two.
Tagalog → English
Tagalog
On the roadmap. Tagalog marks aspect inside the stem, so a lookup has to resolve sumulat back to sulat before it can answer.
Romanian → English
Română
On the roadmap. The reader handles the script today, so this pack needs its dictionary and its cloze bank.
Modern Hebrew → English
עברית
On the roadmap. Right-to-left reading shipped with Arabic, so what is left is the pack itself: an unpointed key for pointed dictionary entries, which is the same fold Arabic uses for its vowel marks.
Danish → English
Dansk
On the roadmap. The Swedish pack proves the pattern, and Danish has more English sentence pairs than Swedish did.
Slovak → English
Slovenčina
On the roadmap. The Czech pack proves the pattern, so this pack needs its dictionary and its cloze bank.
Irish → English
Gaeilge
On the roadmap. A lookup has to undo initial mutation before it can answer, so bhean resolves to bean.
Welsh → English
Cymraeg
On the roadmap. Welsh needs a frequency source of its own, because the word-frequency data every other pack uses covers no Celtic language.
Scottish Gaelic → English
Gàidhlig
On the roadmap. Scottish Gaelic shares its frequency work with Welsh and Irish. The sentence bank stays smaller than most, because little Gaelic and English parallel text carries an open licence.
Biblical Hebrew → English
עברית מקראית
On the roadmap — pending reader support for its script.
The reader works with any language. A pack adds the sentence banks and the dictionaries. Ask for the language you want. We build the packs that the most people ask for.
Which language do you want next?
Choose a language and we will email you once, on the day it is ready. No newsletter.
Your build order comes from this list. Want to talk it over? Join the Discord.