thoughtasylumTOOLBOX Preview Guides Use offline Automate Feedback

Tools › Images

Image to Text (OCR)

Copy the text out of a screenshot, photo, scan or scanned PDF with optical character recognition (Tesseract) in 47 languages, or make a searchable PDF, run in your browser, so nothing is uploaded.

About this tool What it's for, how to use it and an example

What it's for

Get the text out of a picture so you can copy, search or edit it: a screenshot of an error message, a photo of a sign or a receipt, or a scanned letter. The reading (optical character recognition, with Tesseract) happens in your browser, so the picture is never uploaded.

For example, when someone sends a screenshot of a long error, paste the screenshot here and copy the text into a search engine or a bug report; or when a scanned letter arrives as a PDF whose text can’t be selected, open the PDF here to copy its words.

How to use it

Choose, drop or paste a picture, and the text appears in the box, where you can correct it, and Copy copies it. A PDF is read a page at a time, each drawn at 300 dpi first, with a --- Page 2 --- line between pages. The first time, the reader itself (about 3 MB) is loaded, which takes a moment; after that it’s kept, and Save everything for offline on the Offline page saves it ahead of time, with the languages you tick there.

Save searchable PDF saves the PDF or picture with the text found laid invisibly over it, so a PDF reader can search it and select its words, and copying gives the text. A PDF’s own pages are kept as they are; a picture becomes a one-page PDF. The hidden text is what the reader found, so a misread word won’t be found by a search.

Open Languages and tick every language in the text: 47 of them, including French, German, Spanish, Russian, Arabic, Persian, Hindi, Bengali, Tamil, Thai, Indonesian, Japanese, Chinese and Korean, with Japanese and Chinese also in vertical columns. A page in two languages, such as English and French, reads better with both ticked, though each one more makes reading slower. The list shows each one’s size; the first time one is used, it’s downloaded from this site (most are 1 to 3 MB) and then kept on this device, so it works offline afterwards and the list says saved beside it. Only the language files are downloaded: your picture never leaves the browser. Clearing this site’s data in your browser’s settings removes the saved languages.

If the result is jumbled, change the Layout and the picture is read again: Automatic suits pages and screenshots, One block of text a paragraph, A single line one line such as a serial number, and Scattered text labels and signs. The languages and layout are remembered in the browser. The message says how confident the reader is.

Example

Take a screenshot of this paragraph and paste it in. The text box shows the paragraph’s words, line by line, with a confidence of about 95%.

Good to know

Printed text only: handwriting is rarely read well. Text that mixes horizontal and vertical Japanese or Chinese is read one way only, as the language ticked says. Small, blurred, skewed or decorative text gives mistakes (such as 0 for O or rn for m), so check the result, particularly numbers. Columns and tables come out as plain lines in the text box, and its layout isn’t kept there (the searchable PDF keeps the page as it was). A long PDF takes a while, as each page is read in turn.

Guide: Putting a PDF together to send

Languages: English

Tick every language in the text; each one more makes reading a little slower.

Each language is downloaded from this site the first time it's used (the list shows how much), then kept on this device for offline use. Your picture is never sent anywhere.

Built with The third-party code and data this tool uses, and their licences

This tool runs on the following open-source work, copied onto this site and served from it:

PDF.js 6.4.299
Copyright
Copyright 2023 Mozilla Foundation
Licence
Apache License 2.0
Source
github.com/mozilla/pdf.js
Notes
The pdfjs-dist package's legacy/build/pdf.min.mjs and pdf.worker.min.mjs, unchanged, with their licence notice at the top.
OpenJPEG, inside PDF.js
Copyright
Copyright (c) 2002-2014, Universite catholique de Louvain (UCL), Belgium, and the other OpenJPEG authors; wrapper Copyright (c) 2024, Mozilla Foundation
Licence
BSD 2-Clause License
Source
github.com/uclouvain/openjpeg
Notes
PDF.js's JPEG 2000 decoder, from pdfjs-dist's wasm/ folder, unchanged, loaded by its worker when a PDF has such images. The folder's LICENSE.txt has every licence.
JBIG2 decoder from PDFium, inside PDF.js
Copyright
Copyright 2014 The PDFium Authors; wrapper Copyright 2026 Mozilla Foundation (Apache-2.0)
Licence
BSD 3-Clause License
Source
pdfium.googlesource.com/pdfium/
Notes
PDF.js's decoder for JBIG2, the black-and-white image format of many scanned PDFs, from pdfjs-dist's wasm/ folder, unchanged.
qcms, inside PDF.js
Copyright
Copyright (C) 2009-2024 Mozilla Corporation; Copyright (C) 1998-2007 Marti Maria
Licence
MIT License
Source
github.com/FirefoxGraphics/qcms
Notes
PDF.js's colour management, for pages with ICC colour profiles, from pdfjs-dist's wasm/ folder, unchanged.
Tesseract.js, with tesseract.js-core 7.0.0 and Tesseract's trained language data (4.0.0_best_int) 7.0.0
Copyright
The Tesseract.js and Tesseract OCR authors
Licence
Apache License 2.0
Source
github.com/naptha/tesseract.js
Notes
Served unchanged apart from source-map comments; the licence is beside them.

Every page also uses Montserrat (SIL Open Font License 1.1).
Everything the site uses is on Credits and licences.

Private: this tool runs in your browser. Nothing you type, paste or choose leaves this page.

Saved you a few minutes? Say thanks with a coffee.

Something wrong with this tool, or missing from it? Report a bug or suggest a feature.

↑ ↓ move↵ openesc close