What happens when you press Read
The page is first found: rows and columns of the picture that are mostly bright are counted, and the longest bright run in each direction is taken as the sheet of paper. When that sheet fills between about a third and nearly all of the frame, the picture is cropped to it; otherwise the whole frame is kept and the page tells you so.
Then the tilt is measured. The dark pixels of the text are projected onto lines at every angle from −15° to +15°; at the right angle the rows of type line up and the projection is at its most peaked. The page is turned by that angle, uneven light is divided out block by block, and the contrast is stretched. Only then is the cleaned page handed to Tesseract, in its page-layout mode, in a worker in this tab.
What it will not do
- It does not read handwriting, and it does not reproduce tables, fonts or the page layout — it gives you the text.
- It does not open PDFs. Export the pages as images first.
- It does not correct perspective: a page photographed steeply from one side is straightened but stays a trapezium. Shoot from above for the best result.
- Very curved pages (the gutter of a thick book) read worse near the spine.
Questions people ask
Is my scan uploaded?
No. The page is cleaned up on a canvas in this tab and read by Tesseract in a background worker in this tab. The reader's files are downloaded from this site the first time you press Read; the scan itself has nowhere to go.
What does the clean-up do?
Three things, each of which you can switch off. It looks for the page as the large bright area against a darker desk or scanner lid and crops to it. It measures the tilt of the text rows (up to 15° either way) and turns the page level. And it divides out uneven light — a shadow from the phone, a yellowed sheet — then stretches the contrast. The page stays greyscale; the reader does its own thresholding. The page as it was read is shown beside the text, so you can see what was done.
Does it keep columns in order?
The reader is run in its page-layout mode, which finds blocks of text and reads a two-column page down the first column and then down the second. Wrapped lines inside a paragraph are joined back into one paragraph by default; untick “Join lines into paragraphs” to keep every line break as printed.
Can I read a document with several pages?
Yes. Read the first page, then press “Add the next page” and choose the next photo or scan; it is cleaned up and read with the same settings and added under a “--- Page 2 ---” line. The count of pages is shown with the text, and the .docx download starts each page on a new page.
Can it read handwriting?
No. The reader is built for printed English and Hindi. When the confidence across a page is low the page says so, because handwriting comes out mostly wrong.
Can I open a PDF?
Not here. This page opens pictures — JPG, PNG, WebP, TIFF, HEIC and the other common formats. Export the scanned PDF's pages as images first.
Where to take this next
- Image to WordRead the text in a photo or scan and save it as an editable .docx. Text only, said plainly.
- Image to textRead the printed English or Hindi text in a picture, in this tab, and copy or download it.
- Hindi image to textRead printed Hindi, and the English beside it, as Unicode Devanagari ready for Word.
- Screenshot to textRead the text in a screenshot, and lay a chat out as Them: and You: lines.