Question paper generator from PDF
Turn a textbook PDF into an exam paper
Upload the book your class actually studies from. It is read once, split by chapter and indexed — and from then on every question is written from its pages, with a trace back to the passage it came from.
Five papers free. No card.
Read this first
The one PDF that will not work
A scanned book is a picture of text, not text. If your PDF was made by photographing or photocopying pages, there is nothing in the file to read, so there is nothing to write questions from. This is the most common reason an upload disappoints, and it is worth checking before you upload forty files rather than after.
The test takes five seconds: open the PDF and try to select a sentence with your cursor. If the text highlights, it will work. If you can only draw a box around it, it will not.
There is one near-miss worth knowing about. Older Hindi and Sanskrit books — including a number of NCERT PDFs — are typeset in legacy non-Unicode fonts such as Kruti Dev and Walkman-Chanakya. They look correct on screen but extract as Latin gibberish, and most tools index the gibberish. Evegen recognises those fonts by name and converts the text back to proper Unicode Devanagari before doing anything else, so those files do work. The equivalent legacy fonts in other Indian scripts are not converted.
Where text really is unreadable — a scan, or a script that could not be recovered — the file is flagged rather than indexed, so your library does not quietly fill with nonsense that produces nonsense questions later.
What does work: a publisher’s PDF, anything exported from Word or Google Docs, most official board and NCERT downloads, and notes typed by a teacher. If you have no digital book at all, the standard NCERT library can be connected during setup with nothing to upload.
What happens to it
From upload to first question
The text is extracted
Page by page. This is also where an unreadable file is caught — if what comes out does not look like the language it should be, the file is flagged rather than indexed.
Chapters are found
From the book’s own contents page where it has one, from the embedded PDF outline where it has that, and from heading typography — the size and weight of chapter titles against body text — where it has neither. You can correct the result, and supply a contents page yourself for a book that defeats all three.
Passages are indexed
The text is split at paragraph and sentence boundaries into passages of roughly a thousand characters, each overlapping the last slightly so a definition split across a boundary is not lost. Each passage is converted into a searchable numerical form on our own server, by a model installed on that machine — no third party is involved in this step and the passages are not sent anywhere to perform it.
Questions are written from retrieved passages
When you generate a paper, the passages relevant to the chapters you selected are retrieved and sent to the language model with your paper structure. Not the whole book — the relevant excerpts. Each question it writes carries the excerpt it came from, and that trace is checked before you see the paper.
Practical details
The things people ask while uploading
More than one book
Upload as many as you like, including several for one subject. Notes, a prescribed textbook and a question bank can all sit behind the same class and be drawn on together.
Big files
A full textbook is normal. Indexing runs in the background and you are told when it is ready; you do not sit and watch it.
Per-paper extras
Alongside your permanent library, you can attach context documents to a single generation — a syllabus circular, a chapter handout — as PDF or Word. Those apply to that paper only.
Who can see it
Material belongs to your workspace. Another institution cannot see it, search it or generate from it. A new workspace starts private.
Copyright
Most textbooks are copyrighted, and whether you may load a particular title into a tool like this depends on your licence to it. That judgement is yours — it is set out in the terms.
Deleting it
Remove a book and its index goes with it. Close the workspace and the content it held is deleted, as described in the privacy policy.
What it will not do
Uploading a PDF of past question papers is useful, but not the way people expect: it is good material to check repeats against and to derive a paper structure from. It does not let the generator re-ask those questions verbatim, because questions are written from source material rather than copied from a bank.
Questions
Can Evegen generate question papers from a PDF?
Yes — that is the intended way to use it. Upload the textbook, notes or question bank your class studies from, and it is read once, split by chapter and indexed. From then on every question generated for that class is written from passages of that file.
Will it work on a scanned book or a photocopy?
No. A scanned page is a picture of text, not text, so there is nothing to extract and nothing to write questions from. A text-based PDF — the kind you get from a publisher or by exporting from Word — works. Older Hindi and Sanskrit PDFs typeset in legacy non-Unicode fonts are a special case: those are converted back to proper Unicode automatically, so they do work.
Do I have to tell it where the chapters are?
Usually not. Chapters are detected from the book’s own table of contents where it has one, from an embedded PDF outline where it has that, and from heading typography where it has neither. You can supply a contents page yourself if a particular book defeats all three, and you can always correct the detected chapter list afterwards.
Can I upload Word files instead?
The library upload is built around PDF, which is what schools actually have. When you generate a single paper you can also attach extra context documents as PDF or .docx for that paper only.
What happens to my file after I upload it?
It is stored in your workspace and indexed on our own server — the step that converts passages into a searchable form runs on our machine using a locally installed model, and no third party sees it. When you generate a paper, the relevant passages are sent to the language model that writes the questions. Your material is never used to train models and never pooled with another institution’s.
Related
Upload one book and see what it writes.
Five papers free. One textbook is enough to judge it properly.