Get Better Text from Screenshots and Photos

Choose the language in the image

The language of the website controls and the language of the text in a picture are separate choices. Select the writing present in the image. Simplified and Traditional Chinese use separate options. If an image mixes languages, the selected recognition model may handle some characters better than others.

Use the clearest original image

An original screenshot usually gives the recognizer clearer character shapes than a compressed copy. Crop away irrelevant borders while keeping entire letters and lines. Correct a sideways image with the rotation setting. Blurred, very small, slanted or shadowed text can remain difficult even after these adjustments.

Expect plain text, not the original layout

OCR output can be edited, copied or downloaded as TXT. It does not reconstruct a formatted Word document or a spreadsheet table. Columns may be read in an unexpected order. Compare the output with the source rather than assuming every line belongs in the sequence returned.

Review the characters most likely to matter

Check names, reference numbers, decimal points, punctuation and similar shapes such as 0 and O. If you compare two versions with Text Compare, that comparison can show changes but cannot tell which version is correct. Keep the source image while reviewing.

Know what downloads and what stays local

The browser downloads recognition code and the chosen language resources. The image and extracted text are processed on the device. The first run can therefore need more time and bandwidth than a later run using cached resources. Cancelling a task stops recognition; it does not delete the original file from your device.

Continue with a related tool