Image to quiz: from a textbook photo to cited questions
Sometimes the source isn't a file — it's a page in a physical book, a printed worksheet, or a whiteboard at the end of class. This guide covers turning a photo into a cited quiz: how OCR recovers the text, why grounding still applies, and when a photo is (and isn't) good enough.
- ✓OCR quality sets the ceiling — a clear, flat, well-lit photo is worth more than any setting.
- ✓Even from an image, every question should cite the recovered sentence it came from.
- ✓Printed text is reliable; clean handwriting often works; messy handwriting won't.
- ✓Review the citations — OCR mistakes are visible when the source text is shown beside the question.
How image-to-quiz works
QuizRoom runs optical character recognition (OCR) on your image to extract the text, then numbers the recovered sentences exactly as it would for a PDF. From there it's the same grounded pipeline: each question points back at a recovered sentence, and anything it can't cite is rejected. The one added variable is OCR accuracy — which is why the source text is always shown beside the question for you to sanity-check.
An image is just another input — OCR feeds it into the same sentence-indexed pipeline.
Get a clean capture
- ●Flatten the page — hold the book open flat or weigh it down; curled pages distort text near the spine.
- ●Light it evenly — avoid shadows and glare from overhead lights; natural side light is ideal.
- ●Fill the frame — get the text large and square-on rather than at an angle.
- ●One page at a time — a tight crop OCRs better than a wide shot of a whole spread.
- 1Upload the photo
Drop in a photo or screenshot. QuizRoom runs OCR and numbers the extracted sentences.
- 2Generate and check
Questions stream in, each grounded in the recovered text — and the recovered text is shown so you can catch any OCR slip.
- 3Review and deliver
Accept the good items, then export or publish a hosted quiz.
If you already have a digital PDF, use it — born-digital text skips OCR entirely and is more accurate. Photos shine when the material only exists on paper. Either way, see the PDF-to-quiz guide for the file-based flow.
Why grounding rescues OCR
OCR occasionally misreads a character — an “rn” becomes an “m,” a decimal point vanishes. In a summary-based generator, that error would silently propagate into a question you'd never catch. Because QuizRoom shows the recovered sentence next to the question, a bad read is obvious and you can drop or fix the item before it reaches a student. Grounding doesn't just prove the question is faithful — it makes the pipeline's mistakes visible.
What if my photo is a bit messy?
OCR handles most textbook and slide photos. Because the recovered text is shown beside each question, you can review every item and drop anything where the read went wrong.
Does handwriting work?
Clearly printed text works best. Neat handwriting often works; dense or messy handwriting is unreliable — a typed or printed source will always be better.
Can I upload several pages?
Yes — upload multiple images and QuizRoom indexes them together into one source for generation.
Is a photographed page as good as a PDF?
A born-digital PDF is more accurate because it skips OCR. Use photos when the material only exists on paper; use a PDF when you have one.
Turn a photo into a cited quiz
Snap a page and get grounded questions in seconds — with the recovered text shown for every one. Free with 60 credits.
Start free — 60 creditsBuilds the generation pipeline behind QuizRoom, including the citation check that rejects any question it can't trace back to your source. Previously shipped ingestion and OCR systems at two edtech companies.