What OCR is and how to improve accuracy when reading a photo of a document

By Marcelo Amantino · published October 5, 2026 · 2 min read · who we are

OCR (optical character recognition) is the technology that reads the text that appears in an image and gives it back as real text, which you can copy, search and edit. It solves an annoying problem: documents that only exist as a photo or a scanned PDF. But it depends a lot on the quality of the image, and that is where you can improve the result.

How OCR works, in a few words

The program analyzes the image, locates text regions, separates lines and characters and compares each shape with what it knows about the letters and words of the chosen language. That is why indicating the right language improves the result, especially with accents and common words.

On WebFaz, the default mode of OCR runs in your browser, without sending the image. There is also an optional AI mode for handwriting and difficult photos; it is only used if you tick the option, and in that case the image is sent to an outside service, which the page warns you about first.

Why OCR makes mistakes

Almost always the problem is in the image, not in the program:

Shaky, out-of-focus or low-resolution photo.
Uneven lighting, shadows and glare on the paper.
Document tilted or photographed at an angle.
Very small letters, decorative fonts or text over colored backgrounds.
Handwriting, which varies a lot from person to person.

Checklist to photograph better

Place the document on a flat surface and photograph it from above, with the camera parallel to the sheet.
Use natural light or uniform light; avoid direct flash, which creates glare.
Get close enough for the text to fill a good part of the image, but keep everything sharp.
Crop to leave only the area with text.
On colored or faded documents, an image with good contrast helps more than a high-resolution one.

Step by step in WebFaz OCR

Upload the image or the scanned PDF.
Choose the language of the text.
Keep the local mode; consider the optional AI only if the document is not sensitive.
Review the text and export as TXT, Word or Excel.

Always review what cannot be wrong

OCR can swap a 0 for an O, a 1 for an l or omit a comma, and in numbers that changes the meaning. Check values, dates, tax IDs, codes and names carefully. For important documents, compare the text with the original image before using it.

If the PDF already has text

Not every PDF needs OCR. If you can select and copy the text, it is already searchable. OCR is for PDFs that are just images of pages, such as those produced by a scanner. To merge, split or compress PDFs before or after recognition, use the PDF tool.

When OCR is not the best way out

Documents with complex tables, several columns or many graphic elements can come out with the text order mixed up. In those cases, use the result as a draft and reorganize it by hand. For the text already extracted, the Text Editing tool helps clean up lines, remove duplicates and standardize.