OCR — Optical Character Recognition — is the technology that turns a picture of text into real, editable, searchable text. To a computer, a scanned page or a photo of a document is just a grid of coloured pixels. OCR looks at those pixels and works out which letters and words they represent, so the result is text you can select, copy, search and edit.
What OCR means
If you can't select the text in a PDF or image, there is no text there — only an image of text. OCR creates the missing text layer. It's what lets you copy a paragraph from a scanned book, search inside a scanned contract, or turn a photographed receipt into a spreadsheet.
How OCR works, step by step
- Pre-processing: the image is cleaned up — straightened, contrast-adjusted and converted to make characters stand out.
- Layout analysis: the page is split into regions — blocks, columns, lines and individual words — using the geometry of the ink.
- Character recognition: a trained model matches the shapes of letters to characters, using a language model to choose between look-alikes (like “0” vs “O”).
- Output: the recognised characters are assembled back into words and lines, each with a confidence score.
Beyond text: document reconstruction
Basic OCR stops at a flat wall of text. A better approach uses the positions of the words to rebuild the document's structure. Our OCR studio reads the geometry the engine produces and reconstructs headings, paragraphs (repairing words split across line breaks), lists and aligned tables — then lets you export to Word, HTML, Markdown or a searchable PDF. That's the difference between “extracted text” and a document you can actually use.
Why accuracy varies
OCR is probabilistic, so quality of input drives quality of output:
- Resolution: aim for at least 300 DPI; low-resolution scans blur characters.
- Contrast & lighting: dark text on a light background recognises best.
- Skew: crooked pages confuse line detection — straighten first.
- Language: pick the correct language so the right character model is used (English, Hindi, Bengali, Odia, Tamil, Telugu, Marathi, Gujarati, Punjabi and more are supported).
- Handwriting & stylised fonts are far harder than clean printed text.
Confidence scores help here: our tools highlight low-confidence words so you can fix only the uncertain ones.
What OCR is used for
- Making a scanned PDF searchable without changing how it looks.
- Turning a scan into an editable Word document.
- Copying text from a photo or screenshot.
- Digitising receipts, forms, notes and old documents.
FAQ
Is OCR free here? Yes, and it runs entirely in your browser — your document is never uploaded.
Does it work on photos, not just scans? Yes — use Image to Text for photos and screenshots.
Why are some words wrong? Low resolution, skew or poor contrast reduce accuracy. A sharper image and the correct language usually fix it; you can also edit the result directly.
Can OCR read handwriting? Printed text is reliable; handwriting is much harder and results vary.