OCR reads text from the pixels in your photo. If the photo is blurry, tilted, shiny, or cut off, OCR has to guess. So better results start with a better photo and end with you checking the values that matter.

1. Improve the light

Use broad, even light. A window or a lamp with a shade usually works better than a small bright light right above the page. Watch for a white hotspot that crosses a line of text. If the page is glossy, change the phone angle rather than increasing exposure until the paper looks white.

2. Give the camera enough detail

Move close enough that small text is sharp and large on screen, but far enough that all four edges fit. Tap to focus if the camera keeps refocusing. Clean the lens if small text always looks hazy. Digital zoom cannot fix a page photographed from too far away.

3. Correct geometry before recognition

A page photographed at an angle has slanted lines and letters of different sizes. Hold the phone parallel to the page when you can, then use perspective correction to make the page square. If you cannot see a corner clearly, take the photo again instead of forcing a crop that can cut off part of the page without you noticing.

4. Choose a filter for the document

DocumentGood starting pointWatch for
Printed contractColor or neutral grayscaleFaint text removed by very strong black-and-white settings
ReceiptHigh-contrast grayscaleThin decimal points and small totals
HandwritingColor or gentle contrastPen strokes removed as if they were background marks
Stamped formColorColored stamps disappearing in black-and-white

5. Match the language and layout

OCR works better when it knows the page’s language. Pages with more than one language, vertical writing, columns, tables, and unusual fonts need more checking. If the app does not let you choose a language, use the review screen to find the words OCR was unsure about.

6. Review the right things first

Do not spend the same time on every word. Check first the values that change what someone does: names, account numbers, dates, prices, addresses, medication instructions, and legal phrases. Compare them with the original image and correct them before you export.

  • Search for a word from the top, middle, and bottom of each page.
  • Compare every amount and decimal point.
  • Check characters that look alike: 0/O, 1/I/l, 5/S, 8/B.
  • Read the page visually after OCR; the scan image is what counts.
When retaking is faster

If the text is blurred, cut off, or covered by glare, do not keep running OCR again. Taking a clean new photo usually works better than processing a bad photo again and again.

Example

Imagine you keep changing OCR settings, but the photo itself is blurry. OCR cannot read detail that is not in the photo, so better light, focus, crop, and perspective will help more than changing another setting.

Fix things in the same order each time: light and background, framing, rotation, cleanup, language, then OCR. Check the hardest page, and keep the original if a filter might remove something from the page.

Questions people ask

Does higher resolution always improve OCR?

A bigger image helps only if it is sharp. A sharp, well-lit page at a normal resolution is more useful than a huge image with motion blur or glare.

Why does OCR fail on a table?

Tables have small text, lines, and columns. OCR may still read the words but put them in a different order from the one on the page.

Can OCR read handwriting?

Some engines support handwriting, but results vary widely with pen style, spacing, contrast, and language. Use OCR on handwriting to save time, but check what it reads against the page.