Copy the text out of a picture
A screenshot of an error message, a photo of a printed letter, a scanned contract, a slide in a meeting, a receipt: the words are right there, but you cannot select them. Typing them out again is slow and easy to get wrong. This tool reads the text in an image or a scanned PDF (optical character recognition, or OCR) and gives it back as text you can copy, edit and paste anywhere.
Many online OCR services upload your image to a server. This one does not: the recognition engine runs inside your browser, so private documents stay on your computer.
How to use it
- Choose an image or a PDF, or simply copy a screenshot and paste it on this page with Ctrl+V.
- Wait a few seconds. The first time, the recognition engine (about 7 MB) is downloaded from this site; after that it is kept by your browser and starts quickly.
- Check the text and correct any misread words, then copy it or download it as a text file.
Getting the best results
- Sharp and straight: hold the phone parallel to the page, fill the frame with the text, and avoid shadows. Blurry or tilted photos are the main cause of mistakes.
- Big enough: text should be at least 20 pixels tall in the image. Small images are enlarged automatically, but detail that is not there cannot be added.
- Printed text works best. Handwriting is rarely recognised reliably by any free OCR.
- Simple layouts: one column reads best. For tables, the words come out in reading order, but the columns are not kept; check them carefully.
Good to know
- Language: the tool is set up for English. Other languages that use the same Latin letters (French, Spanish, German and others) mostly work, though accents may be missed.
- Scanned PDFs: each page is read in turn. PDFs that already contain real text do not need OCR; get their text with the PDF text extractor instead, which is instant and exact.
- Common mix-ups to check by eye: the letter O and the number 0, the letter l and the number 1, rn read as m, and full stops lost after small numbers. Numbers in invoices and account details deserve a second look before you rely on them.
- Accuracy is shown as the engine's own confidence. Above 90% usually means very few mistakes; below 70%, check every line.
How this tool was tested
Before publishing, I ran 20 automated checks on the finished page in Microsoft Edge, which is built on Chromium like Chrome. Last full run: 15 November 2026.
- Real recognition, served by this site: the actual Tesseract engine and English model were loaded from this site (no other server) and read test images with known text: printed Arial text came back at least 95% word for word, italic serif text at least 90%, and small 13-pixel screenshot text, enlarged first, at least 85%.
- The sample notice was read correctly, line breaks could be kept or joined, and hyphenated words at line ends were mended.
- Scanned PDFs: an image-only PDF page was read through pdf.js and OCR.
- Using the tool: the confidence figure, a blank image giving no text, the .txt download and a file that is not an image.
What the tests did not cover: Handwriting, photos of crumpled or curved pages and languages other than English were not tested. Safari and Firefox were not tested. Something not working? Tell me on the contact page.
FAQ
Is it free?
Yes, with no sign-up, no page limit and no upload.
Can it read text from a screenshot of a video call or slide?
Yes, as long as the text is sharp. Paste the screenshot with Ctrl+V.
Does it work offline?
After the first use, the engine is stored by your browser, so it keeps working without a connection while the page is open.