How to Extract Text From an Image
How to extract text from an image using OCR: what affects accuracy, how to prepare images for best results, and what to expect from different source types.
To extract text from an image, upload it to the OCR tool — it reads the image and outputs editable text in seconds. Nothing is uploaded to a server; everything runs in your browser.
How well it works depends on the image. Clear printed text comes back near-perfect. A blurry photo of handwritten notes in poor lighting will not produce usable results. The rest of this post explains how to get good results and what to watch for.
What Works and What Doesn't
OCR works reliably on:
- Printed documents, typed letters, and books
- Scanned receipts and invoices
- Business cards
- Screenshots — on-screen text is pixel-sharp, so accuracy is consistently high
OCR is less reliable for:
- Handwriting — especially cursive. Neat block printing gives the best chance, but expect some errors.
- Very small text — footnotes or fine print photographed from normal distance
- Text over a patterned or busy background
- Images taken in poor light or at an angle
How to Get the Best Results
Get close. The text should fill most of the frame. Shooting from arm's length with the document in the corner of the shot loses detail.
Use even lighting. Natural window light or a lamp positioned directly above the page works well. Shadows across text reduce accuracy.
Shoot straight down. Hold the camera directly above the page and parallel to it. Tilted text produces more errors.
If scanning: use 300 DPI or higher. Lower settings reduce character detail enough to affect accuracy.
Fix contrast before uploading if needed. If the image looks washed out or shadowy, run it through your phone's photo editor first. High-contrast black text on a white background is what OCR reads best.
How to Extract Text From an Image: Step by Step
- Prepare the image. Crop to just the text area. Fix contrast and straighten if needed. If your phone has a document scanner, use it — it handles contrast and straightening automatically.
- Upload to the OCR tool. On mobile, you can also use the camera directly — convenient for business cards, signs, and menus.
- Copy the output. Review for obvious errors before using. Printed documents in good condition come back clean.
Cleaning Up Extracted Text
Even clean output is worth a quick check. Watch for:
Hyphenated words. Older documents break words across lines ("encoun-\ntered"). Rejoin them manually.
Missing paragraph spacing. Paragraphs may run together or pick up extra line breaks. Reformat to match the original.
Tables. Table contents come out as a flat list — left to right, top to bottom. Rebuild the structure manually.
Multi-column layouts. A two-column page may be read across columns instead of down each one. Check the reading order before using the output.
Keep Sensitive Documents Private
The OCR tool processes images locally — your file never leaves your device. For medical records and legal contracts, this matters. Tools that upload to a server hand your document to a third party's infrastructure. Local processing eliminates that.
If you need to compare two versions of a document before pulling the text — checking what changed in a contract revision, for example — the PDF comparison tool shows page-level differences with the same on-device approach.