Extract text from an image

OCR, optical character recognition, turns the letters in a picture into text you can copy, search and edit. Use this page for a photographed notice, a screenshot of an error message, a scanned letter saved as an image, or a slide from a presentation.

The recognition engine is Tesseract, served from FileGizmo’s own site and started only when you ask for text. The first run downloads the engine and its English language data; after that the browser can keep them, so later images start faster and the tool keeps working offline. The image stays in the current tab from start to finish.

OCR is an interpretation of the picture, not a copy of it. Clear, upright printed text at a readable size gives the best results. Low contrast, glare, handwriting, decorative fonts, text over photos, skewed angles and very small type all make mistakes more likely. The result shows an average confidence as a guide, but a confident result can still misread a name, a date or an amount.

Before you use the text, compare anything that matters with the image. Numbers are worth checking twice, because an OCR engine can read a 5 as an S or a 1 as an l without any warning. Tables and columns come out as plain lines of text, so their layout usually needs tidying by hand.

For better results, take the photo straight on and in even light, crop away anything that is not the text, and make sure the letters are in focus. A screenshot is usually recognized better than a photo of a screen, because it has no glare and no angle.

This release reads English. It may manage other languages written in the Latin alphabet, but it is not tuned for them, and FileGizmo does not claim it reads them accurately. If your image is part of a scanned PDF, the OCR PDF tool reads the PDF directly and can add the text to it as a searchable layer.