How InPage stores Urdu
An InPage file holds its text in InPage’s own encoding, not in Unicode: each letter is a pair of bytes, and the letters run in the Urdu alphabet’s order from ا onwards. The text is a chain of paragraphs, each one its length followed by its bytes, and text boxes such as a title page or footnotes are chains of their own. The converter reads that chain from the file itself and writes each letter and diacritic as Unicode, so the result can be searched, copied and edited in Word.
InPage keeps a number in the order it is drawn, right to left, so 1427 is stored as ۷۲۴۱. And its honorific font puts صلی اللہ علیہ وسلم, رضی اللہ عنہ and the rest on single keys, so the file holds a Latin e or t standing alone in an Urdu sentence. The converter turns digits round and writes each honorific out as words, read from the name it follows, and the report counts how many it wrote.
What other conversions get wrong
A conversion that copies the codes as they are writes every number backwards, the way the sample shows, and leaves the honorifics as stray Latin letters: the Unicode Word files that came with the books checked here print رسول اللہ e where the page shows the blessing after the Prophet’s name, and two diacritics as private-use characters that most fonts cannot draw. A date, a page reference or a verse number written backwards reads as a different number and nothing on the page says so.
Measured
The table was read from 25 real InPage files from public Urdu book collections, typed from the 1990s to 2019, aligned with the Unicode Word files their publishers made of seven of the books. Against those files, once their own choices are allowed for (honorifics left as Latin letters, private characters, the elongation stroke dropped), 99.94% of 140,511 characters agree. The table was learned from the same files, so that agreement is not independent evidence; it says the reading is consistent, not that it has been checked against a fresh book.
On the live site on 26 October 2026, two chapters of one of those books converted in 3 and 8 seconds. The Word files hold 99.07% and 99.54% of the characters in the publisher’s own Unicode file, counted without spaces, and every difference was read. Most are the publisher’s choices: honorifics the converter writes as words and the publisher left as Latin letters, two diacritics the publisher stored as private-use characters, the elongation stroke it dropped, and ؤ where the typist typed ئو. The footnotes, which the publisher’s file keeps apart, follow their text in ours. The rest are 11 of the 12 unknown signs described below, where the publisher’s file has nothing. Both Word files are set right to left in Noto Nastaliq Urdu. No Urdu reader has checked a converted book against its printed pages yet.
Where it still goes wrong
Only the text and its paragraphs come across. InPage’s fonts, sizes, bold, alignment, page layout and pictures are not read yet, so the Word file is set in one size of Noto Nastaliq Urdu. InPage 3 files are not read at all: no real one has been found to measure, and the converter says so rather than guessing.
About a dozen signs that InPage draws from its symbol fonts, such as bullets and a year sign, have no known Unicode equivalent yet. They come out as �, and the report counts them so they can be checked against the printed page; one of the two chapters above had 12. In the PDF made beside the Word file each one is drawn as a black diamond holding a question mark, so it stands out on the page. Footnote numbers are not stored in the file, because InPage numbers footnotes itself, so each footnote mark is written as *.