Principles of the Transliteration (Add-On Version 3.7+)
AynAyn offers transliteration into Hebrew and Latin. In Options → Translit → Semitic scripts, Translit target selects the target for Hebrew, Arabic, Syriac, Phoenician, and Imperial Aramaic text. Hebrew is the default; Latin requires explicit selection. Ethiopic has its own Ethiopic → Latin checkbox, which is off by default and works with either target.
The Other scripts subtab has individual, default-on Latin transliteration checkboxes for Devanagari, Bengali/Assamese, Gurmukhi, Gujarati, Odia, Tamil, Telugu, Kannada, Malayalam, Sinhala, Armenian, Georgian, Greek, Thai, and Japanese. These settings are independent of the Semitic target. Unchecking a script preserves its original letters. Translit mode is on the main Translit tab. Suppress comments and its Omit silent wasla and article letters suboption are in Semitic scripts.
Hebrew spelling controls for dagesh, Hamza vowels, Shva, and final-letter forms are grayed out and disabled with the Latin target. Smart mode is available with Latin. Suppress comments, Shadda duplication, standalone Madda interpretation, and Ancient Hejazi reconstruction are available for Hebrew and Smart Latin, and disabled for pure ISO Latin. Disabled controls retain their saved settings when switching targets or Smart mode.
Sections: Hebrew transliteration · Latin transliteration · Ethiopic vowels · Ge’ez numerals · Display modes and direction.
Hebrew transliteration
Aramaic and Arabic are close relatives of Hebrew, and transliteration into the Hebrew writing system makes many words recognizable to a Hebrew speaker. The transliteration model is designed to make Arabic words easily recognizable to a Hebrew reader; pronunciation follows Literary Arabic.
For Arabic words, the converter reads letters together with their vowel marks and uses the word form and surrounding text where the spelling is ambiguous. It recognizes the definite article with attached conjunctions or prepositions, but also distinguishes a similar beginning that belongs to the word itself. A silent article lam before a sun letter may appear in square brackets, as in (א[ל]שַّמס) الشَّمْس. When the article is clear, the converter can restore an unwritten shadda on the sun letter.
In Hebrew, the letters בגדכפת have, or historically had, two variants of pronunciation: plosive and fricative. Dagesh Qal denotes a plosive variant. The lesser-known overline called rafe may denote a fricative one. The converter uses both signs. It also uses geresh to mark a letter whose sound differs from its Hebrew pronunciation when both variants are plosive or both are fricative. Where possible, the converter does not use modern Hebrew geresh-based Arabic transliteration. In modern Israeli Hebrew, only ב, כ and פ still have two pronunciations. In Liturgical Mizrahi Hebrew, all בכפת letters have two pronunciations; ג and ד retain two pronunciations in Yemeni Hebrew. The converter uses Liturgical Mizrahi and Yemeni pronunciations of בגדכפת to reflect specific Arabic sounds in transliterated text.
In Arabic, פ/ف always represents F, ב/ب always represents B, and כ/ك always represents K. However, د ط ح ص ت, the equivalents of דטחצת, each belong to a pair of sounds, like Hebrew בגדכפת. The variant whose pronunciation differs from the basic Hebrew or Aramaic pronunciation—ordinarily a fricative—is marked with an additional dot that mostly acts like rafe. Although ح has the same sound as Mizrahi Hebrew ח, while خ has the same sound as כֿ, the converter maps خ to ח׳. Earlier versions used the phonetically consistent כֿ, but خ is linked etymologically with ח, not כֿ. It is thought that two Proto-Semitic sounds that remain distinct in Arabic merged into one Hebrew sound and letter. Conversely, one Arabic or Proto-Semitic sound represented by ك is thought to correspond to two distinct Hebrew sounds.
With dagesh and without rafe, דכת in transliterated text denote the same sounds as in modern Israeli Hebrew; ט represents the emphatic /tˤ/ of Arabic and some traditional Mizrahi Hebrew pronunciations. By default, Arabic plosive דּ and תּ retain dagesh. The Omit dagesh from ungeminated ד and ת option removes this plosive mark, including from Ta Marbuta rendered as ת; rafe continues to mark דֿ and תֿ. Dagesh is retained for written or restored Shadda, including when the geminated consonant is duplicated. Syriac Qushshaya always maps to Hebrew dagesh regardless of this option. In modern Israeli Hebrew, צ is pronounced as an affricate /ts/, which differs from both Arabic ṣād /sˤ/, a voiceless emphatic fricative, and ḍād /dˤ/, a voiced emphatic plosive. In the Tiberian reading tradition and Yemenite Hebrew, however, צ has an emphatic fricative pronunciation /sˤ/, similar to Arabic ṣād. To distinguish the two Arabic sounds explicitly, the converter uses צֿ with rafe for ص (ṣād), and צׄ with a dot above for ض (ḍād). The dot follows traditional Judeo-Arabic notation and makes the letter easier to read than the previously used dagesh. Within this transliteration system, the two signs distinguish the fricative and plosive sounds. The latter is pronounced similarly to a voiced alveolar plosive /d/, but with additional pharyngealization. For example, مصر (Egypt) is transliterated as מצֿר, while أرض (earth) is transliterated by default as ארץׄ, preserving recognizable Hebrew spelling. With rafe and without dagesh, the דטת letters denote the following sounds, as described in the IPA guide to Arabic:
| דֿ/ذ | the “th” sound in the English word “this” |
| טֿ/ظ | the “th” sound in the English word “thus” |
| תֿ/ث | the “th” sound in the English word “thin” |
The letter غ GHAYN denotes an R-like sound, the voiced uvular fricative. Many Ashkenazi Jews pronounce ר this way. It would be phonetically consistent to transliterate غ as רֿ, but the converter maps it to ע׳ because غ is linked etymologically with ע, not ר. In converted Syriac text, softened gamal is rendered as גֿ; for example, ܒܓ݂ܕܕ → בגֿדד. A special dot “רוככא” marks a “soft” letter in the original text.
The letters ر/ר represent the same sound. In Arabic, it is a trill close to the Amharic or Russian R. This sound is very difficult for most Ashkenazi Jews to pronounce. During the Holocaust, local Nazi collaborators in Eastern Europe used the Russian word “KUKURUZA”/“кукуруза” as a shibboleth to identify Jews for extermination.
The letter and diacritic sign ء HAMZA , whether written independently or on a carrier, denotes a glottal stop /ʔ/, while ordinary Arabic alif without HAMZA often represents a long /aː/ vowel. The converter represents HAMZA by Hebrew alef א, which historically denotes the same consonant. In pointed Hebrew, an alef carrying its own vowel point normally represents a consonant rather than a vowel letter. The converter follows this principle: an alef bearing an ordinary vowel point, shva, or a hataf is treated as consonantal. A word-initial alef is also consonantal even without pointing. A shva belonging to the immediately preceding consonant is retained where appropriate, allowing HAMZA's alef to carry its following vowel directly.
Alif can also preserve an older orthographic notation of a glottal stop. In traditional spellings such as مِائَة (one hundred), this retained alif occurs alongside the modern seated HAMZA denoting the same glottal stop. The duplication is orthographic, not phonetic: the alif itself is now ignored. The Add-On omits it without square brackets and handles HAMZA normally; for example, مِائَة → מִ[י]אֲ|תּ with hataf-patah enabled. This handling applies to a bare alif, with neither HAMZA nor its own vowel except optional sukun, after kasra before HAMZA on yeh, or after damma before HAMZA on waw. Missing vowels in recognized hundred-word forms are recovered only where unambiguous, with one practical exception: unpointed مائتان and مائتين are preferentially read as the numeral 200. Explicit conflicting vowel marks take precedence; other ambiguous forms such as مائتي are not covered by this preference, and competing morphological analyses are retained. The same omission is applied before Ancient Hijazi reading reconstruction: the retained alif is not treated as a long /aː/ vowel, and the following HAMZA is then processed according to that mode's reconstruction rules. For the historical orthography, see Al-Jallad (2021), §3.2.
The Add-On represents HAMZA together with its following short vowel using either an ordinary Hebrew vowel point or a hataf, according to the selected options. This applies both to vowels explicitly written in the Arabic text and to unwritten vowels that can be reliably inferred from HAMZA's position, carrier, and context. The choice of notation depends on the vowel being represented, rather than merely on the Arabic letter carrying HAMZA. Hebrew hatafs traditionally represent reduced vowels and occur primarily under guttural consonants. In this transliteration, however, they are used as an alternative notation for HAMZA followed by an ordinary short Arabic vowel, without implying any additional vowel reduction. For short /a/ (Arabic fatha), the available notations are hataf-patah אֲ and ordinary patah אַ. For short /i/ (Arabic kasra), the available notations are ordinary hiriq אִ, synthetic hataf-hiriq אְִ, and hataf-segol אֱ. Synthetic hataf-hiriq combines shva and hiriq on the same alef. Although this is not a traditional Hebrew hataf, it expresses the intended /i/ vowel while retaining the visual shva component. Hataf-segol is another optional approximation, although the corresponding Hebrew vowel differs from Arabic /i/. For example, إِ → אְִ when synthetic hataf-hiriq is selected. For short /u/ (Arabic damma), ordinary qubuts אֻ provides the closer phonetic correspondence. Hataf-qamatz אֳ can be selected as an alternative. In Hebrew, hataf-qamatz represents a reduced /o/-like vowel, historically [ɔ̆], rather than /u/. Although both are rounded back vowels, they are not identical. Traditional Hebrew has no hataf-qubuts sign.
When a corresponding hataf option is disabled, the Add-On uses the ordinary Hebrew vowel point directly beneath HAMZA's alef: patah for short /a/, hiriq for short /i/, and qubuts for short /u/. An alef carrying its own vowel point is already consonantal, so no additional alef with shva is inserted merely to carry an explicitly written vowel. If the preceding consonant has sukun, its corresponding Hebrew shva is retained on that consonant, and HAMZA takes its following short vowel directly, regardless of the hataf settings.
Hatafs can represent only short vowels. Long vowels following HAMZA are therefore written using the ordinary Hebrew notation for long vowels, regardless of the hataf settings: qamats for /aː/, hiriq male for /iː/, and shuruk for /uː/. The same consonantal alef carries qamats or hiriq and represents the glottal stop: medial /ʔiː/ is written אִי. Before shuruk, the vowel point belongs to waw rather than alef, so a medial consonantal alef retains its shva: אְוּ. At the beginning of a word, this shva is omitted because the initial alef is already consonantal.
When HAMZA's following vowel is not explicitly written, the converter attempts to recover it from HAMZA's position, carrier, and linguistic context. Although these provide useful clues, they do not always determine the vowel unambiguously. Where a short vowel can be reliably recovered, it is represented just like an explicitly written vowel, using the selected hataf or ordinary vowel point. When the vowel cannot be determined, a consonantal alef within or at the end of a word retains the Arabic HAMZA sign above it: אٔ. This distinguishes it from a vowel-lengthening alef without implying that no vowel follows. Shva is retained where sukun establishes the absence of a vowel, including at the end of a word. For example, سأل → סאٔל and شيء → שיאٔ. At the beginning of a word, neither this HAMZA sign nor a consonantal shva is added, since the initial alef is already unambiguously consonantal. For example, when the initial vowel cannot be determined, أغلب → אע׳לבּ. This omission does not affect shva when it forms part of synthetic hataf-hiriq אְִ. When the initial vowel is known, the corresponding ordinary vowel point or hataf is placed directly on the initial alef. No additional alef with shva is inserted before it. For example, إيمان → אִימָאן when ordinary hiriq is selected. Within a word, however, if an unwritten kasra is represented by ordinary hiriq, a separate alef may carry that vowel after אְ where necessary to preserve the explicit consonantal marker. The omission of redundant initial shva does not remove a hataf or the shva belonging to a preceding consonant.
Meteg מֶתֶג, also called gaʿya גַּעְיָה, is a traditional Hebrew reading sign: a small vertical stroke below the letter, usually to the left of its vowel point, as in אָֽ. In pointed Tanakh Hebrew, it can mark secondary stress and indicate that a vowel should receive a fuller or lengthened pronunciation. Its precise function depends on the reading tradition and phonological context. See Gesenius, §16. AynAyn uses this association with sustained pronunciation to represent the vowel elongation indicated by Arabic Madda. This is a transliteration convention, not an assertion that Hebrew meteg and Arabic Madda have identical historical functions.
Alif Madda آ denotes /ʔaː/ and is written אָֽ: one consonantal alef with qamats and meteg. No additional alef with shva is inserted. A preceding consonant's shva is retained where appropriate; for example, قُرْآن → קֻרְאָֽן.
Standalone Qur'anic Madda marks ٓ and ۤ indicate vowel elongation. The converter represents this indication by meteg next to the corresponding Hebrew vowel, including hiriq male and shuruk, without adding another HAMZA. Meteg follows the vowel when its written Arabic carrier differs from the Hebrew vowel's carrier, and repeated marks for the same vowel produce one meteg. By default, a separately typed alif + Madda is treated as elongation; turning off Read standalone Madda as vowel elongation permits an ambiguous combination to be read as Alif Madda. Alif after written fatha remains a vowel-lengthening carrier.
For example, the modern Arabic spelling إِسْرَائِيل is transliterated as אְִסרָ[י]אִיל when synthetic hataf-hiriq is selected. The Qur'anic spelling إِسْرَٰٓءِيلَ is transliterated as אְִסרָֽאִילַ under the same setting. In both forms, the initial HAMZA with short /i/ is represented by synthetic hataf-hiriq, while the medial HAMZA before long /iː/ requires ordinary hiriq male on the same consonantal alef. Meteg preserves the written Madda's indication of elongation. Square-bracketed י or ו associated with HAMZA are annotations of its source spelling carrier.
The Add-On also provides an Ancient Hijazi reading reconstruction mode, based on research into the early Qur'anic consonantal text and its historical pronunciation. In this mode, HAMZA is generally omitted in positions where it is reconstructed as absent from Ancient Hijazi pronunciation, and the corresponding historical vowel sequences or long vowels are reconstructed instead. Written Madda's meteg is retained with the resulting vowel.
ى ʾalif maqṣūrah (ألف مقصورة) denotes a long /aː/ vowel at the end of a word in Literary Arabic. It looks like a dotless ـي. The converter marks the long vowel with kamatz and may show the silent source letter in square brackets, as in عَلَى → עַלָ[י]. Egyptians often omit the dots from final ي, making it look like alif maqsurah. The converter uses vowel marks, recognized word forms, and related forms in the same text to distinguish the two readings. When the evidence identifies an undotted final yeh, it renders Hebrew י; for example, فِى → פִֿי. Recovery is not applied when the reading remains ambiguous.
The shva in converted Arabic text should not be pronounced as an additional vowel. When it corresponds to sukun (سكون), it denotes the absence of a following vowel. An unknown vowel after noninitial HAMZA is instead indicated by the Arabic HAMZA sign above Hebrew alef, אٔ. Kamatz denotes long /aː/ in converted Arabic and Eastern Syriac texts, but /o/ in converted Western Syriac.
Qur'anic initials are read by letter names when their characteristic marks identify them; for example, الٓمٓ → אֲלִףֿ לָֽאם מִֽים. Unmarked الم is also read by letter names before ذلك when preceded by the full basmala or followed by the first-verse marker. Elsewhere, the unmarked form remains ambiguous.
Latin transliteration
Smart mode, beside Translit target, is checked by default and is available when Latin is selected. It applies the shared Arabic/Syriac reading rules. Unchecking it selects pure ISO conversion from the written source, retaining encoded vowel marks and vowel letters without reading transformations or Ancient Hejazi reconstruction.
| Script | Standard or reference |
|---|---|
| Hebrew | ISO 259:1984 |
| Arabic | ISO 233:1984 |
| Indic scripts | ISO 15919:2001 |
| Armenian | ISO 9985:2026 |
| Georgian | ISO 9984:2026 |
| Greek | ISO 843:1997, Type 1 |
| Thai | Based on ISO 11940-2:2007 |
| Japanese | kuromoji-es with bundled IPADIC dictionaries, followed by WanaKana romaji conversion |
| Ethiopic | ALA-LC Amharic (2011), including Ge’ez, with additional forms from ALA-LC Tigrinya (2011) |
The ISO notation is a representation of writing. For example, Hebrew dagesh uses an overdot, as in בּ → ḃ; qamats is å and shewa is °. Arabic bare alif is ʾ, independent hamza is ˌ, and hamza on a carrier is ˈ. Shadda uses an overline: بَّ → b̄a. These conventions differ from the pronunciation-oriented Hebrew notation described above.
Additional Arabic-based letters used in Persian, Urdu, Pashto, Sindhi, Kurdish, and Uyghur extend the ISO 233 mapping for shared Arabic letters. For example, پ چ ژ گ → p č ž g. Syriac, Phoenician, and Imperial Aramaic use corresponding Semitic Latin consonants as AynAyn extensions; ISO 259 and ISO 233 do not define those mappings. Syriac written vowels and consonantal dots are represented where supported.
Indic transliteration covers the ten scripts listed above. It retains encoded inherent vowels, uses virama to suppress the inherent vowel, and handles vowel signs and consonant combinations. It does not guess language-specific schwa deletion, so an ISO written form can differ from the modern spoken word. Greek uses Type 1 transliteration, rather than the Type 2 pronunciation rules; for example, η ω β → ī ō v. Armenian and Georgian use the listed 2026 editions.
Thai transcription uses a local pronunciation dictionary and approximate spelling rules for unresolved words. Tone and vowel length are not marked. When a spelling has several dictionary readings, the converter takes the first available reading; context-sensitive pronunciation may differ. Ethiopic pronunciation options are also approximations, as explained below.
Japanese transliteration is enabled by default and works with either Semitic target. Kuromoji supplies dictionary readings for kanji; WanaKana converts those readings and kana to romaji. Written long vowels are retained, for example 東京 → toukyou and コーヒー → koohii. The particles は, へ, and を use wa, e, and o when identified as particles. Original spaces and punctuation are preserved, and halfwidth or decomposed kana are normalized only for conversion. Kanji without a dictionary reading remain unchanged. Names and ambiguous kanji may have other readings; Han-only text is interpreted using the Japanese dictionary while this option is enabled. The dictionaries load only when needed and work offline. Library licenses and dictionary notices are included in the browser packages.
Japanese readings are analyzed in sentence context before being mapped to the original clickable words. This preserves suffix and verb readings, for example 女性史 → joseishi, 管理下 → kanrika, and 持たず → motazu, as well as doubled consonants in あった → atta. Word lookups retain the sentence context. When a compound reading cannot be divided safely across the original word boundaries, its source spelling is retained.
Japanese romanization adds spaces between words while keeping suffixes and inflected verb forms together, for example chuugoku no joseishi de wa and atta. Combined mode also separates word pairs. Original-only mode retains the source spacing.
In combined mode, words containing kanji show their contextual hiragana reading before the romanization: 食べる (たべる, taberu) and 学校 (がっこう, gakkou). Kana-only originals keep their usual original-and-romanization display. The hiragana reading is clickable and looks up the original word in context. Transliteration-only mode shows romanization without the added kana annotation.
Arabic and Hebrew presentation forms are expanded before conversion. For enabled source scripts, letters and signs without a defined Latin equivalent are skipped, including unsupported annotations. Text from unrelated scripts is preserved. Transliteration runs locally in the browser; online dictionary and reference searches are separate.
In Smart Latin, Arabic pronounced hamza is ʾ (U+02BE), and confirmed long vowels are ā ī ū; their lengthening alif, waw, or ya is absorbed into the vowel. A bracketed alif remains [ʾ]. Annotated hamza seats include the ISO hamza mark, for example [wˈ]ʾ and [yˈ]ʾ. ISO hamza marks ˈ and ˌ occur only inside spelling annotations in smart output. Standalone madda uses a circumflex on its vowel. Arabic sukun is used in reading analysis but omitted from both Smart Latin variants, including detached marks. Pure ISO retains its ° notation; Hebrew shewa and literal degree symbols are unchanged. Other letters, shadda, and tanwin retain their Latin notation. These Arabic-specific alif rules do not remove Syriac alaph.
Suppress comments selects Smart uncommented output: square brackets and their silent contents disappear. With comments suppressed, silent wasla alif and assimilated article lam are omitted by default in Hebrew and Smart Latin. The Omit silent wasla and article letters checkbox is checked by default; uncheck it to retain both letters without brackets. It is available only with comments suppressed and does not change pure ISO output. Braces around reconstructed pronounced sounds disappear while the sounds remain, so {ʾ} becomes ʾ. Source punctuation is preserved. Pure ISO output is independent of this checkbox.
Ancient Hejazi reading reconstruction is available for Hebrew and Smart Latin, and disabled for pure ISO Latin without clearing its saved value. Both smart renderers share the Arabic reconstruction, including long ē/ō, hamza loss, glides, and g. For example, عَلَى gives ordinary Smart Latin ʿalā[ỳ] or reconstructed ʿalē[ỳ]. The reconstruction does not alter Syriac.
Ethiopic vowels
Enable Ethiopic → Latin to transliterate supported Ethiopic letters. The base consonants and syllables follow the ALA-LC Amharic table, whose scope includes Ge’ez, with additional Tigrinya forms. The seven ALA-LC vowel orders are a u i ā é e o. Two subcontrols are checked by default and become available when Ethiopic letter transliteration is enabled:
- Approximate sixth-order vowels controls whether the optional sixth-order vowel is retained. The existing spelling rules keep a word-initial vowel, including in a one-letter word; omit a final vowel in a longer word; and retain a medial vowel when the preceding syllable has no vowel. The vowel-carrier rows እ ዕ and the labialized rows keep their explicit table vowels. Unchecking this control retains every sixth-order vowel.
- Pronunciation-oriented vowels controls how the retained vowels are written. First-order vowels use e, except for the historical guttural rows ሀ ሐ ኀ አ ዐ and the labialized ኈ row, which keep a. Retained sixth-order vowels use ɨ. This control changes only vowel representation; it does not change the omission rules. Unchecking it restores ALA-LC vowel spelling while keeping the selected omission behavior.
For example, with both subcontrols checked, ለ → le, ሀ → ha, ሰላም → selām, and ግዕዝ → gɨʻɨz. For ሰላም, unchecking only Pronunciation-oriented vowels gives salām; unchecking only Approximate sixth-order vowels gives selāmɨ. Unchecking both gives the full ALA-LC written form salāme.
The pronunciation-oriented vowel spelling follows the Amharic distinctions described in Appen’s language-specific documentation, using e for the first-order central vowel and ɨ for the retained sixth-order vowel. Fourth-order ā, other vowel orders, and explicit wa/ya forms retain their table spelling. These options do not resolve every Ethiopic language’s pronunciation or infer unwritten consonant doubling.
Ge’ez numerals
Ge’ez numeral conversion is always enabled in transliterated output, independently of the Semitic target, the Ethiopic letter checkbox, and both vowel subcontrols. Traditional numeral sequences are interpreted as complete numbers using additive coefficients and the hundred and myriad multipliers. For example, ፪፻፱ represents 2 × 100 + 9 = 209, and ፲፰፻፹፪ represents (10 + 8) × 100 + 80 + 2 = 1882. Large values are converted without rounding. See the published Ge’ez numeral examples.
As a compatibility convention, a complete numeral run made only of the unit signs ፩–፱ is read digit by digit. Thus ፩፰፰፪ becomes 1882. Sequences containing tens, hundreds, or myriads retain the traditional arithmetic; ፲፱, for example, becomes 19.
Combined view displays an original numeral followed by its decimal value, such as ፪፻፱ (209). A numeral inside a word stays joined to its letter parts: በ፩፰፰፪ becomes በ1882 with Ethiopic letter transliteration off, or be1882 with it on and the default vowel settings. Combined view shows the whole pair, በ፩፰፰፪ (be1882). Clicking either representation looks up the complete original word.
Display modes and direction
Translit mode offers original text only, transliterated text only, or original text with transliteration. Original-only mode keeps the source spelling, including its numeral signs. Combined mode preserves the original language’s word order and paragraph direction: Arabic, Hebrew, Syriac, Phoenician, and Imperial Aramaic remain ordered from right to left. Each word transliterated into Latin is displayed from left to right within that order.
With the Latin target, transliteration-only mode uses left-to-right paragraph and word order. Individual Latin words from the independently enabled scripts also read left to right when the Semitic target is Hebrew. The Ayn Ayn context-menu preview follows the selected target and script options. Changing display preferences updates the visible transliteration while retaining the source text used for dictionary searches and word lookups. Ornate Arabic parentheses around verse numbers keep facing inward in every display mode, including Latin transliteration-only output; each number stays together with its brackets and remains clickable.