How to Extract Text From an Image With OCR

Optical character recognition turns pictures of words into words. It is the difference between a photo of a page and a document you can search, copy and edit. Modern OCR is remarkably good — as long as the image gives it something to work with.

What makes OCR succeed or fail

Recognition accuracy depends almost entirely on the input:

• Resolution. Aim for text at least 20 pixels tall. A tighter crop beats a distant wide shot.
• Contrast. Dark text on a light background. Avoid photographing through plastic wallets.
• Angle. Hold the camera parallel to the page. Perspective distortion confuses line detection.
• Focus and shadow. Move so your own shadow is not across the page.
• Language. Tell the tool which language to expect; it changes the character model.

Step-by-step: OCR in Docsy

1. Open the OCR tool.
2. Upload your photo, screenshot or scan.
3. Select the document language.
4. Start recognition — progress runs through loading, recognising and finishing stages so you can see it working.
5. Review the extracted text, then copy it or download it as a file.

If the tool reports no readable text, the image almost always needs a retake rather than a retry: crop tighter, add light, and hold the camera straight.

Proofreading the output

Even excellent OCR needs a skim. Check these first:

• Digits: 0/O and 1/l are the classic confusions — critical in invoice numbers and IBANs.
• Line breaks in justified columns.
• Tables, which arrive as plain lines of text and need rebuilding.
• Accented characters, if you chose the wrong language.

Going further

Recognised text can go straight into an AI workflow: summarise a scanned report, translate a foreign-language notice, or pull the key dates out of a letter. Once the words are real text, everything else becomes possible.

FAQ

What image quality do I need for good OCR?

Sharp, evenly lit, and cropped to the page. Text should be at least 20 pixels tall — around 300 DPI for a normal printed page.

Does OCR work with handwriting?

Printed text is far more reliable. Neat block handwriting sometimes works; cursive generally does not.

Which languages are supported?

Major Latin-script languages are supported, and selecting the correct one before recognition noticeably improves accuracy.

What if no text is found?

That means the image had no recognisable characters — usually too small, too dark or too angled. Retake the photo closer and straighter.

Try it on a photo

Take a picture of any printed page and see how much of it comes back as editable text.