Recognize text from the screen
Use the Screen region OCR command to draw a rectangle on the screen and capture text that cannot be copied directly from an app, browser, remote session, or PDF viewer.
How HyperSnap OCR works, and how to tune it.
HyperSnap uses the open-source Tesseract OCR engine to recognize editable text from selected screen regions, screenshots, and image files when normal copy and paste is not available.
What OCR does inside HyperSnap
Recognize text from an image
Use Image or selection OCR to read the current image tab, or limit recognition to a selected rectangle inside the image.
Send OCR output where you need it
Recognized text can be placed on the clipboard, shown in Notepad, or saved to a text file depending on your OCR setup choices.
Recommended setup steps
1. Review OCR Setup first
Open the OCR Setup dialog before you begin. That is where you choose output behavior, OCR languages, and the trained data set you want HyperSnap to use.
OCR is powerful, but it is not error-free. Small text, compressed images, low-contrast UI, or icons that resemble letters can still produce mistakes.
2. Pick the right language mix
Set the correct primary OCR language first. If the captured text mixes two languages, add a secondary language so Tesseract has the right character set and dictionary hints.
Choose the OCR data set
Tesseract offers three trained-data families. The best choice depends on your language, image quality, and whether you prefer speed or maximum accuracy.
Fast
Best when you want responsive OCR and your source text is already fairly clean. This is often a good default for English screen text.
Regular
A middle ground between speed and quality. Use this when Fast is not accurate enough, but Best feels slower than necessary.
Best
The most accuracy-focused option, and usually the slowest. Try it when the other sets struggle with your language or image quality.
More information about the upstream OCR data sets is available from the Tesseract tessdata project. Tesseract itself is distributed under the Apache License 2.0.