NovaWhat OCR does, and what it doesn't change
OCR adds a layer of recognized text positioned exactly over the words it read, invisible on top of the existing page image. The page still looks like a scan — nothing about its appearance changes — but a reader can now search for a word and jump to it, select a sentence, or copy text out instead of retyping it by hand.
It doesn't fix a bad scan — if the source image is too blurry or low-resolution for the text to be legible to begin with, OCR can't recognize what a human eye can't read either. The output quality directly follows the input quality.
What affects recognition accuracy
Clean, printed text on a plain background at a reasonable resolution recognizes best — the kind of scan a flatbed scanner or a well-lit phone photo produces. Handwriting, low-contrast text, skewed pages, and heavily stylized fonts all reduce accuracy, sometimes substantially.
If accuracy on a specific document matters a lot, a quick manual check of a page or two after running OCR is worth it — search for a word you know is on the page and confirm it's found, rather than assuming every page recognized perfectly.
Languages and longer documents
Picking the right language before running OCR matters — recognition is tuned per language, and running English recognition over a French or Spanish document produces noticeably worse results than selecting the correct language first. OCR runs one page at a time, so a longer document takes proportionally longer than a short one; there's no way to meaningfully speed that up beyond picking the correct language the first time so you don't need to redo it.