Why You Can't Copy Text from a Photo and How OCR Reads It

Contents
Photos and screenshots store text as pixels, not letters. Drag your mouse across the words and nothing highlights. That's because an image file is just a grid of color values — it holds no text data at all. OCR (optical character recognition) analyzes the shapes of those pixels and turns them back into actual characters, and how well it works depends heavily on the quality of the photo.
Summary ① Photos and screenshots store letters as pixels, not text, so you can't select them by dragging. ② OCR compares the shape of each letter against learned character patterns and picks the closest match. ③ Recognition rates swing widely with image quality, contrast, tilt, and whether the text is printed — handwriting does noticeably worse.
Why You Can't Select Text in a Photo
A photo of a meeting-room whiteboard, a scan of a paper document, and a screenshot of a messenger conversation all share the same structure. The file stores a color value for every x/y coordinate — and nothing that says "this letter is here."
To your browser, an <img> tag is one solid picture. With a text document, the browser recognizes each individual character and paints the dragged area blue. An image has no "characters" to paint in the first place.
So Here's How to Check: Starting With the Most Reliable Method
The most reliable method is a person looking at the photo and typing it out. It's the most accurate, but even a single page takes several minutes, and longer documents invite typos.
OCR automates that process. It slices the image into small pieces, compares the stroke shape of each piece against character patterns it learned in advance, and guesses the closest match.
As the name "recognition" suggests, it doesn't know the right answer — it picks the most probable character. That's why bad conditions can make it read one letter as another.
Check it with Image to Text (OCR). Instead of retyping every sentence, drop in a photo file and it reads Korean and English characters right in your browser, turning them into editable text. Your image is never sent to a server — everything happens inside your browser.
Free · No sign-up · Your image is never uploaded

The 4 Factors That Decide Recognition Accuracy
Unlike a person, OCR doesn't understand context. It only looks at the shape of the pixels. So even with the same tool, results vary a lot depending on the state of the photo.
| Factor | Good conditions for recognition | Bad conditions for recognition |
|---|---|---|
| Resolution | Sharp enough that letter strokes are clearly visible | Low resolution or blur that smears the strokes together |
| Contrast | Strong contrast, like black text on a white background | Weak contrast from a patterned background or pale colors |
| Tilt | Lines of text sit level with the frame | Shot at an angle, creating perspective distortion |
| Letterform | Printed type (a fixed font) | Handwriting, cursive, or decorative fonts |
Printed type produces the same letters in nearly the same shape every time, so it matches the learned patterns well. Handwriting is the opposite: stroke thickness, slant, and the way letters connect differ from person to person, so pattern matching gets far less accurate.
The Bottom Line
Text not being selectable in a photo isn't a broken tool — it's how image files are built. If you need an exact sentence, typing it yourself is the surest route, but for long documents it's faster to let OCR produce a draft and have a person review it. To improve accuracy, the most important thing is to use sharp, high-contrast photos of printed text shot straight on and level.
Free · No sign-up · Your image is never uploaded
Frequently Asked Questions
Can it read handwritten notes?
It's optimized for printed type, so it works best on printed text. Handwriting varies so much from person to person that accuracy can drop sharply — for a short note, retyping it yourself may actually be faster.
How should I take the photo to get better results?
Hold steady and make sure it's in focus, and avoid glare or shadows covering the text. Shooting the document straight on and level reduces tilt, which noticeably improves recognition.
Is my image uploaded to a server?
No. Text recognition runs entirely in your browser (on your own device), and your image is never sent out. One caveat: the first time you run it, it downloads the Korean and English language data once, so it may be a little slow to start.
Can it read photos that mix Korean and English?
Yes. It's set up to recognize Korean and English together, so you don't need to shoot them separately even when both appear in the same document.
This article was adapted from our Korean original. Read the Korean original