Extract text from a photographed page
Improve OCR results by preparing the image and checking the copied text.
Text recognition reads shapes in an image and guesses the characters they represent. It is convenient for turning a printed notice into editable text, but it cannot guarantee that a bank number, name or deadline was read correctly. Treat the result as a draft to review.
Give the recogniser a clean image
Use a sharp photo taken straight above the page. Fill the frame with the text while leaving a small margin so characters at the edges are not cut off. Avoid glare, folds, shadows and patterned backgrounds. If the page is sideways, use Image Cropper to rotate it by 90 degrees. If the page is tilted at another angle, take another photo before trying Image to Text.
What to expect from one page
Imagine a printed notice that says “Applications close on 30 September.” If the photo is blurry, “30” might be read as “80,” and a narrow letter might disappear. The tool extracts what it can see; it does not know the notice's intended deadline. After recognition, place the image and extracted text side by side. Check dates, amounts, names, punctuation and line breaks before copying the text into another document.
When recognition is poor
- Try a brighter, higher-resolution photo with the entire line in focus.
- Crop away unrelated borders or illustrations so the text occupies more of the image.
- Break a large page into sections if tiny text becomes unreadable on a phone.
- For a PDF with selectable text, use PDF to Text instead. Copying the text layer is usually more accurate than reading a screenshot.
Handwriting, unusual fonts and mixed languages can lower accuracy. Keep the original image alongside the output when the transcription will be used for an important decision.