How to extract text from an image
- 1Choose the languages of the text. English is preselected, or the language of your browser.
- 2Drop images or a scanned PDF on the box, tap it to pick a photo, or paste a screenshot with Ctrl + V (Cmd + V on a Mac).
- 3Check the text in the box below each image, fix what the OCR got wrong, then copy it or download everything as a .txt file.
What OCR is and how it works here
OCR, optical character recognition, turns a picture of letters into real text. This tool uses Tesseract, the open source OCR engine that Google developed for many years, compiled to run in your browser. It finds the lines of text in the image, and a neural network trained on each language reads them word by word. That is why the right language matters: it knows which letters, accents and words to expect.
Small images such as screenshots are enlarged before reading, because the engine works best when letters are around 30 pixels tall. PDF pages are rendered at about 300 dpi, the resolution scanners use for text.
Tips for better results
- Screenshots beat photos. Take a screenshot of text on a screen instead of photographing the screen.
- Flat, sharp and bright. For paper, lay the page flat, hold the phone straight above it and avoid shadows and glare.
- Crop to the text. Logos, photos and patterns next to the text can turn into stray characters. The image cropper helps.
- Clean up the result. Text from narrow columns often comes with line breaks in the middle of sentences. Remove line breaks joins them, and the word counter shows how long the text is.
Want the subject of a photo instead of its text? The background remover cuts out people, pets and products, also right in your browser.
Frequently asked questions
Are my images uploaded?
No. The text recognition runs in your browser with the open source Tesseract engine. Screenshots of chats, scanned contracts and photos of documents stay on your device.
Which languages can it read?
English, German, French, Spanish, Italian and Portuguese, including accents and umlauts. Select every language that appears in the image: with two languages selected, a page that mixes them is read correctly. Each language downloads its data once, between 0.7 and 3 MB.
How accurate is it?
For screenshots and clean scans of printed text, nearly every word comes out right. Photos of documents work well when the page is flat, sharp and evenly lit. Handwriting, decorative fonts, very small print and text on busy backgrounds give poor results, so check the text before you use it.
Can I get the text out of a scanned PDF?
Yes. Drop the PDF and each page is rendered at about 300 dpi and read like an image, up to 50 pages at a time. If the PDF already contains real text that you can select, the PDF to Word converter keeps its headings and layout.
How do I get the best results?
Use the sharpest version of the image you have, crop away anything that is not text, and hold the phone straight above the page when you take a photo. Screenshots work better than photos of a screen. Choose only the languages that are really in the image.