Skip to content
Home / Tools / Image to Text (OCR)

Image to Text (OCR)

Upload, drop or paste one or more images and extract the text with Tesseract.js. Choose languages, layout, rotation and enhancements. Nothing is uploaded to Fakulti.

The first run downloads the OCR engine and language data.
Advertisement
728 × 90Ad placeholder · tool-below-workbench

How the OCR works

Optical character recognition turns pictures of text into text you can edit. This tool runs Tesseract.js, a JavaScript and WebAssembly build of the open-source Tesseract engine, entirely in your browser. Your images are not uploaded to Fakulti. The first run downloads the recognition engine and the language data (a few megabytes per language), which your browser then caches.

Prepare the image for better accuracy

  • Resolution: aim for roughly 300 DPI, or text that is at least 20–30 pixels tall. Tick “Upscale small images” for low-resolution screenshots.
  • Contrast: dark text on a plain light background works best. Use “Grayscale” and “Boost contrast” for faded scans or photos taken in dim light.
  • Straighten and crop: a skewed photo or a page edge with shadows lowers accuracy. Crop to the text before uploading.
  • Right language: choosing the correct language matters more than any other setting. Add a second language for mixed text, such as English with Malay.

Layout mode (page segmentation)

Automatic suits normal pages. Single block works for one column of text such as a paragraph screenshot. Single line is for a cropped line like a caption or a licence plate number. Sparse text finds scattered words, useful for posters, signs and diagrams. Switching modes is the quickest fix when output looks scrambled.

What OCR cannot do well

Handwriting, decorative fonts, very small print, heavily compressed JPEGs, curved text and complex tables remain difficult. Multi-column layouts may be read straight across the columns. Always proofread numbers, names and anything legal or financial; characters such as 0/O, 1/l/I and 5/S are classic mix-ups.

After extraction you can join broken lines, tidy spacing, copy the text or download it as .txt. The last result is kept in your browser so a refresh does not lose it.

Enhancement controls

If the first result is poor, adjust the image before recognition. Rotate fixes sideways scans, Grayscale and Boost contrast help faded print, Black and white with a threshold slider turns uneven backgrounds into clean text on white, Invert handles light text on dark backgrounds and Upscale enlarges small screenshots. Use Preview processed image to see exactly what the engine will read before you run it. After extraction, the confidence figure for each image tells you which results deserve a closer proofread, and the tidy buttons join broken lines, collapse spaces or flatten the text to a single line.

Advertisement
728 × 90Ad placeholder · tool-in-article

Frequently asked questions

Is my image uploaded to a server?

No. Recognition runs locally. Only the engine and language files are downloaded from a public CDN.

Why is the first run slow?

The engine and language data must be downloaded once. Later runs are faster because the files are cached.

Can it read handwriting?

Not reliably. Tesseract is designed for printed text; neat block capitals sometimes work but expect many errors.

Can I process several images?

Yes. Select multiple files and they are processed in order, with the results combined in one output box.

What does the confidence figure mean?

It is the engine’s own estimate of how sure it is about the recognised words. Below about 80% it is worth proofreading carefully.

When should I invert the image?

When the text is light on a dark background, such as a dark-mode screenshot. Engines expect dark text on light.

Related reading