Scanned Image to Text

In short A scanned document is a picture of a page, not a document. There is no text inside it at all, which is why searching finds nothing, copying selects nothing and a screen reader has nothing to announce — the file is a photograph that happens to contain writing. Recognition is what converts that picture back into characters. Export or screenshot the page as an image and this reads it, entirely in your browser, which matters given that scans are usually contracts, records, medical letters and other things nobody should be uploading to a stranger's server. Scans behave differently from screenshots and reward a little preparation. Scan at 300 DPI rather than 150 — below about 200 the character shapes start to degrade in a way no amount of processing recovers. Straighten the page, because even a two-degree skew measurably hurts accuracy across a full line of text. Prefer greyscale or black and white over colour, and if the original is a photocopy of a photocopy, expect to proofread. One page at a time is the honest workflow here: this reads an image, so a long multi-page PDF means exporting the pages you actually need rather than the whole file.