You have a scanned contract, a photo of a printed page or a PDF that will not let you select any text. To get the words out, you need optical character recognition, usually shortened to OCR. It looks at the shapes in the image and works out which letters they are.
Check whether you need OCR
Open the PDF and try to select a word with your cursor or finger. If you can, the PDF already contains real text, and you can copy it directly. DocPro’s Extract text tool reads that text straight from the file. If the whole page highlights as one block, or nothing happens, it is a picture and needs OCR.
Get a good scan first
OCR accuracy depends mostly on the picture. A few minutes spent on the scan saves a lot of correcting later.
- Lay the page flat. Curved pages near a book spine distort letters.
- Use even, bright light. Avoid shadows across the text and glare from glossy paper.
- Hold the phone square to the page. Tilted photos squash letters and make lines slope.
- Fill the frame. The text should be large and sharp. Around 300 dpi is a common target for scanners.
- Choose high contrast. Dark text on pale paper works best. Faint pencil, coloured backgrounds and stamps over text cause errors.
- Straighten and crop. If the picture is rotated or has a wide border, fix that in Edit image first.
Run OCR in DocPro
- Open Extract text and choose an image or a PDF.
- Pick the language or languages that appear in the document. The tool supports English, Arabic, Persian, Urdu and Hindi. Choose every language that is actually on the page, because more languages make it slower.
- For a PDF, choose which pages to read.
- Run the extraction, then copy the text or download it as a .txt file.
Recognition runs in your browser and the language data is bundled with the page, so nothing is uploaded and nothing extra is downloaded. Extract text is a Pro tool on the web: you can set up your file and languages for free, and the run asks you to sign in to DocPro Pro.
What to expect
On a clean, printed page in a common language, OCR is usually good, but never perfect. Expect small errors, and expect more of them on poor scans, small print, unusual fonts and photos taken at an angle. Handwriting is generally not reliable. Tables and multi-column pages may come out as lines of text in an order you did not intend, and the layout is not preserved.
Proofread it
Always read the result against the original. Typical mistakes are lookalike characters: the letter O and the number 0, a lowercase l and the number 1, rn read as m, and broken accents or diacritics. Pay extra attention to numbers, names, dates, reference codes and amounts, since a wrong digit is easy to miss. Search for patterns, such as double spaces or stray symbols, that point to a bad area.
After extracting
- Paste the text into your word processor and format it. If you want the original layout instead, try PDF to Word on PDFs that already contain real text.
- If your scan is a set of photos, combine them into one document with JPG to PDF.
- If you only have a PDF and need images of the pages first, use PDF to JPG.
Privacy
Scanned documents are often sensitive: IDs, medical letters, contracts. Because OCR here runs on your device, the picture is never sent to a server. Close the tab when you are done, and delete the text file if you do not need it.
OCR output is a draft. Check names, numbers and dates against the original before you rely on them.
A clean scan, the right language and a careful proofread are what turn a picture of a page into text you can use.