Gujarati subtitle & caption generator

Gujarati script comes from the same Nagari family as Devanagari, but its letters don't hang from a headline. In Hindi, the shirorekha along the top of each word binds it into one shape; in Gujarati, each letter stands on its own and only the spaces mark where a word ends. At caption size that puts more weight on the typeface itself: a light Gujarati font has nothing tying its letters together and can break apart against a busy background, where a heavier face stays readable.

Vowel signs still sit above and below the letters, as in ગુજરાતી, so a line of Gujarati needs more height than a line of English at the same size. Styles that stack two tight lines, or draw a background box sized for Latin text, can clip those marks. Give Gujarati captions a little more line spacing than you would give English.

The heavy Latin display fonts popular for short-video captions contain no Gujarati letters at all. Left alone, a device fills the gap with whatever Gujarati font it happens to have, so the style stops applying to the Gujarati words. AutoCap avoids that by giving every caption style a Gujarati fallback, Baloo Bhai 2, and by offering typefaces drawn for the script:

AutoCap can write the same Gujarati speech three ways: in Gujarati script, romanised into Latin letters, or translated into English. The right one depends on who watches. Gujarati script reads naturally to people who learned to read it; romanised text suits viewers who speak Gujarati but read English day to day; an English translation reaches people who don't speak Gujarati at all.

Romanised Gujarati has no fixed spelling, so the same word can fairly appear in more than one form. Choose one spelling for anything you repeat, such as a name or a catchphrase, and correct the transcript to match. When the transcript is in Gujarati script, the editor lists Gujarati fonts first.

Automatic transcription is weakest on words a model rarely hears: family names, place names, brand names, and English words spoken inside Gujarati sentences. Read those first. Numbers deserve a look too, because a spoken number can be written out in words, in Western digits or in Gujarati digits such as ૧૨, and a video reads more cleanly when it sticks to one form.