Switching

Image to Text: Extract Text from Images

Reads the text in photos, screenshots, and scanned documents and turns it into text you can copy. Supports Korean, English, and Japanese. Images are processed in your browser and never uploaded.

Text language

Pick every language that appears in the image. Each extra language adds to the first download and slows recognition.

Try this when text on a colored background is not being read.

Choose images

Drag and drop multiple files at once

Try with an example

How to use

  1. Pick the language of the text. For a document that mixes Korean and English, pick both.
  2. Drag and drop images or click to choose them. You can add several at once.
  3. Press Extract text. The first run takes a little longer while the engine and language data download.
  4. Check the result against the image, fix any mistakes, then copy it or download a .txt file.

FAQ

How accurate is it?
It depends on the image. Sharp printed documents and screenshots are usually read almost word for word, but blurry or tilted photos, uneven lighting, and small text lead to more mistakes. The average confidence is shown above the result with a warning when it is low, so double-check numbers and names against the original.
Can it read handwriting?
Not well. The engine it uses (Tesseract) is trained on printed text, so handwriting mostly comes out as the wrong characters. Vertical text is not supported either.
Are my images uploaded?
No. The recognition engine and language data are downloaded from this site once, and all recognition then happens in your browser. The downloaded data stays in your browser cache, so later runs start right away.
How many languages should I pick?
Only the ones actually in the image. If a Korean document contains English words or email addresses, pick Korean and English together. Picking languages that are not there slows recognition and can add stray characters.

Related tools