Skip to content

OCR

What OCR is, how it works, where it's used, and common mistakes to avoid when making screenshots and scans searchable.

OCR (optical character recognition) converts images of text into machine-readable characters. The input can be a scan, a photo or a screenshot. The output is plain text, often with the position of each word, line and block on the page.

Modern OCR engines use neural network models trained on many fonts and languages. Widely used engines include Tesseract, an open-source project, and Apple's Vision framework, which apps on macOS and iOS can call. Recent macOS versions also offer Live Text, which lets you select text in images in Photos, Preview and other apps.

When OCR output is placed as an invisible layer over the page images of a PDF, the PDF becomes searchable while it looks unchanged.

How OCR works

  1. Preprocess the image: adjust contrast, straighten, remove noise.
  2. Analyze layout: find text blocks, columns, lines and words.
  3. Recognize characters and words with a trained model.
  4. Apply language models to correct likely errors.
  5. Output text with coordinates.

Where OCR is used

  • Making scanned documents and screenshot PDFs searchable.
  • Copying text from images, slides and video frames.
  • Reading receipts, invoices and forms into other systems.
  • Accessibility, where a screen reader needs text instead of pictures.

OCR for screenshots

Screenshots are good OCR input: text is straight, sharp and high in contrast. Accuracy drops with very small text, low-contrast colors, decorative fonts, mathematical notation and text inside diagrams. Retina captures and lossless PNG give the engine more detail than downscaled JPGs.

Temposhot runs OCR when it exports a PDF from the results pane, and everything runs on your Mac.

Common mistakes with OCR

  • Running OCR on heavily compressed or downscaled images.
  • Trusting OCR output for numbers and names without checking.
  • Expecting equations, tables and footnotes to come out in the right structure.
  • Mixing languages on one page without checking the result.

Turning pages into a PDF?

Export a searchable PDF from Temposhot, see save screenshots as PDF on Mac, or convert existing images with Screenshot to Searchable PDF.

Common questions

Is OCR text exact?

No. Accuracy is high on clear text but not perfect. Check important details against the image.

Can I get text out of one screenshot?

Yes. Extract Text from Image reads the text in the browser.

Sources

Taking the same screenshots again and again?

Temposhot captures a screen, window or area on a timer and exports the set as a PDF, images, ZIP or MP4. The free browser tools fix single images.

Download for MacSee features

Related resources

Related terms

All glossary terms