Transliterate
Transliterate. Runs entirely on your device, nothing is uploaded.
Runs on your device. The file is never uploaded.
Transliterate maps accented Latin, basic Cyrillic and basic Greek onto plain ASCII letters. NFKD comes first, combining marks are dropped, and a table of roughly 70 characters is applied after. A capital whose mapping runs longer than a letter capitalises only its first, giving Zh. CJK, Arabic and Hebrew have no table and pass through.
Questions
Which scripts does it convert?
Latin with diacritics, basic Cyrillic and basic Greek. It first normalises to NFKD and strips combining marks, which handles accented Latin, then applies a table of about 70 characters: the 33 Russian letters, 25 Greek letters including final sigma, and eight extra Latin letters, the sharp s, ae, oe, the slashed o, the barred d, thorn, eth and the barred l. Upper-case forms are mapped too.
What happens to Chinese, Japanese, Arabic or Hebrew?
It passes through unchanged. There is no table for those scripts, and the tool does not guess. This is a dependency-free character mapping, not a full ICU transliterator, so if you need Pinyin, romaji or a proper Arabic romanisation this is not the tool. Nothing is dropped or corrupted; you get the same characters back.
How is this different from Remove diacritics?
It does more. Remove diacritics decomposes and strips marks and stops there, so Cyrillic and Greek come out unchanged and the sharp s stays as it is. Transliterate performs that same strip and then maps the characters that do not decompose, so the sharp s becomes ss and a Cyrillic word becomes Latin letters. Use Remove diacritics when you want the original script kept.
Which romanisation standard does the Cyrillic follow?
None formally. The table is a plain single-mapping scheme close to common English usage: zh, kh, ts, ch, sh, shch for the compound sounds, y for the short i and for the hard vowel, yu and ya for the final two letters, and the hard and soft signs mapped to nothing. It is built to give readable, URL-safe Latin text, not to satisfy a library cataloguing standard.
Are multi-letter results capitalised sensibly?
Yes. An upper-case source letter whose mapping is more than one character gets only its first letter capitalised, so an upper-case Cyrillic zh becomes Zh and not ZH. That keeps names looking like names. There are no options on this tool, so the behaviour is the same on every run.