English to Urdu, and Urdu to English
Both directions work the same way. A Word file is edited where it stands: its paragraph styles, tables, lists, header, footer and page settings stay the file’s own and only the words change. A PDF with selectable text is first turned into a Word file and then translated, so the result is a Word file you can edit.
Translated Urdu is set in Noto Nastaliq Urdu, and each paragraph is marked right to left, so Word opens it right-aligned and reading from the right. If that font is not on your computer, choose any Urdu font in Word. Translated into English, the paragraphs read left to right again. Numbers, dates, section references, amounts, e-mail addresses and web addresses are taken out before a sentence is sent and put back afterwards, and they stay in the digits you wrote them in. A sentence that cannot be translated stays in the source language and is highlighted yellow; it is never filled in with a guess.
What was measured for Urdu
On the live site on 26 October 2026, the test notice was translated from English into Urdu, and that Urdu back into English. In both directions 6 of 6 protected items came back exactly: the section reference, the year, the date, the amount, the e-mail address and the web address. Both results were marked Publish-ready. The Urdu file carried right-to-left paragraphs in Noto Nastaliq Urdu, and the English read back close to the original, with “For help” become “For assistance”.
Then a real Urdu document: two pages of a book on Hajj, converted to Unicode from the InPage file it was typed in, and translated into English. 22 of 25 passages were translated; the other three, references to Arabic books, stayed in the original and were highlighted. The tool marked the result Draft, not Publish-ready, which is right: the Urdu prose read sensibly in English, but the Arabic quotations the book sets inside its Urdu were where the translation went wrong.
What to read carefully in an Urdu translation
Arabic inside Urdu. Religious and legal Urdu often quotes Arabic and then gives it in Urdu. In the Hajj pages the Urdu came out well, but each Arabic quotation was translated again on its own, loosely, and in one the quoted words did not come back at all: the English read “The Messenger of God said: ))” with nothing in between. The tool now refuses an answer that has lost words like that: run again after the change, that quotation and one other came back kept in Arabic and highlighted, with the book references, five highlighted passages in all. Read every quotation against the source all the same.
Honorifics. In the same pages a short honorific after a companion’s name was once given as the blessing for the Prophet. Elsewhere names and honorifics were kept, as “(may Allah be pleased with him)” and “(PBUH)”. No Urdu reader has checked these translations yet, so read the result, and have anything official or religious checked by someone who reads Urdu.
Urdu in a scan or an InPage file
A scanned PDF has no text to translate, so it is refused rather than guessed at. Urdu scans are read by IndicPDF Smart OCR, in beta, with a free account. Nastaliq print is hard to read: on 21 proofread pages from seven Urdu books printed between 1858 and the 1930s, the best reading still got 16.7% of characters wrong, so correct the text before you translate it.
Most Urdu books and papers of the last thirty years were typed in InPage, whose files open in nothing else. The InPage converter turns an .inp file into a Unicode Urdu Word file, and that Word file can be translated straight away. Footnotes, endnotes and comments are not translated yet, and text inside a picture is not translated.