Image to Text (OCR): How to Get Accurate Results from Photos and Scans
Optical character recognition (OCR) reads the letters in a picture and turns them into text you can edit. It works well on clean printed text and poorly on blurry photos. These tips make the difference between a clean result and a page of errors.
What affects accuracy
- Sharpness and size: the letters need enough pixels. A scan at about 300 DPI is a good target, and text in a tiny screenshot may need a larger source image.
- Contrast and lighting: dark text on a light, even background reads best. Shadows and glare cause errors.
- Straightness: a photo taken at an angle bends the lines of text.
- Language: choosing the wrong language produces strange characters.
- Handwriting: OCR is built for printed text and handles handwriting poorly.
Steps for a clean result
- Crop the picture so only the text remains.
- Choose the language of the text in the Text language option.
- Press Extract text and wait. The first run downloads language data, so it is slower.
- Read through the result and fix mistakes in the box. Look closely at numbers, because 0 and O, or 1 and l, are easy to confuse.
- Copy the text, or download it as a TXT or Word file.
Which tool should you use?
- Image to Text when you only need the words.
- JPG to Word or PNG to Word when you want an editable document.
- JPG to Excel when the picture is a table.
- PDF to Text for scanned PDFs. It reads real text directly and uses OCR only for pages that are pictures.
Taking a better photo of a page
Lay the page flat, use daylight or an evenly lit room, hold the phone parallel to the paper, and fill the frame with the page. A scanner app that flattens the page and boosts contrast helps even more.
Privacy
The text is recognized in your browser, so your pages, receipts and notes are not uploaded. The only thing downloaded is the language data used for recognition.
Frequently asked questions
How accurate is OCR?
On clean, printed text it is usually very good. On blurry, tilted or handwritten text it drops quickly, so always proofread the result.
Can OCR read handwriting?
Not reliably. It is built for printed text.
Does it work for Urdu and Arabic?
Yes. Choose the language before extracting. Accuracy depends on the font and the quality of the image.
Why is the first run slow?
The first time you use text recognition, your browser downloads the language data. Later runs are faster.
Try the tools
Image to Text · PDF to Text · JPG to Word · JPG to Excel · Word Counter
More guides
How to Remove the Background from an Image (and When It Will Not Work) · How to Create a QR Code That Scans Every Time · Word vs PDF: Which Format Should You Send and How to Convert Safely

