OCR SnipScreenshots and scans
Printed text on a plain background — a screenshot of a chat, an error message, a page from a scanner, a slide — reads at very high accuracy. This is the everyday case and it just works.
A screenshot, a photo of a page, a scanned document, a chart with numbers locked in a picture. Paste it with Ctrl+V or drop the file — the text comes back in seconds, and the image never leaves this computer.
Low score — the text may be in the other language. Switch to "English + Russian" and try again.
The recognition engine — Tesseract, compiled to WebAssembly — and the language files are downloaded from this site once and cached. That is the only traffic, and it happens before you choose anything.
Your image is opened from memory and, if it is small, enlarged first: small type reads far better at two or three times its size.
Recognition runs in a background thread on your machine. The text and a confidence score come back; the image is discarded when you close the tab.
Printed text on a plain background — a screenshot of a chat, an error message, a page from a scanner, a slide — reads at very high accuracy. This is the everyday case and it just works.
A phone photo of a receipt or a page works when it is reasonably straight and lit. Strong angles, shadows across the text and motion blur cost accuracy; the confidence score tells you when to retake.
Poor. No OCR engine reads cursive well, and one that claims to is selling something. Block capitals sometimes work; a signature never will.
Really. Open your browser's developer tools on the Network tab and paste an image: no request carries it. The engine and the language files are fetched from this site — that is a download to you, not an upload from you — and after the first visit they are served from the cache.
Because the screenshot is already in your clipboard the moment you take it. Press Ctrl+V (⌘+V on a Mac) anywhere on this page and it is read straight from there. There is also a page built around exactly that, with the capture shortcuts for Windows and Mac.
English and Russian today. Auto mode reads in English and switches to Russian when the result looks unsure, so you rarely have to choose. More languages are on the way — each one is a file served from this site, not a call to a service.
Yes, this tool is free, with no limit on the number of images. The paid pass on this site is for the document converters — bank statements to Excel and the tools that follow it.
This page gives you the text as lines. A table converter that keeps rows and columns is the next tool on this site; until then, the lines paste cleanly into a spreadsheet if the source was aligned.