OCR: how a picture of a page becomes text
A scan is a photograph
A scanned page looks like a document, but to a computer it is only pixels. OCR finds the shapes of characters in those pixels and outputs the letters they most likely represent.
What decides accuracy
Resolution, contrast and straightness matter most. A flat page, even light, a plain font and a resolution around 300 DPI give the best chance. Shadows, creases, tilt and tiny print each cost accuracy.
Phone photos of documents
A phone can work well. Lay the page on a table, hold the camera parallel above it, turn on the room light rather than the flash, and make sure the whole page is in focus.
Always proofread
Even a good result can swap similar characters, such as 0 and O, or 1 and l. If the text contains amounts, dates, phone numbers or names, check them against the original.
OCR or PDF to Text
Use OCR when the file is a scan or photo. Use PDF to Text when you can already highlight the words, because reading an existing text layer is faster and more exact than recognising pictures.
If results look poor, rescan before anything else. A straighter, sharper picture fixes more errors than any other change.