Image to text (OCR)
Pull the words out of a photo, a scanned page or a screenshot, ready to copy, edit or save.
Processed on our server, then deleted
Drop an image here
JPG, PNG or WebP. Or drag it here, or paste a screenshot (Ctrl+V).
Sent to our server for this one job, then deleted. Never stored or logged.
One image at a time: JPG, PNG, WebP and other formats your browser opens, up to 40 MB to start with (it is prepared on your device, then sent at up to 10 MB). English text. Processed on our server, then deleted. 10 free server jobs a day per person, shared by all server tools. More: the OCR API (paid per call).
How to get text from an image
- Choose an image, drag it onto the page, or paste a screenshot with Ctrl+V (Cmd+V on a Mac).
- Check the preview: the text should be the right way up.
- Press Extract text, then copy the text or save it as a .txt file.
From picture to text
A photo of a page is just coloured dots to a computer. OCR, optical character recognition, finds the shapes of letters in those dots and turns them back into text you can search, edit and paste. This tool sends one prepared copy of your image to our server, where the Tesseract engine reads it, and shows you the words in seconds.
Tips for better results
- Hold the phone straight above the page, in good light, and fill the frame with the text.
- Crop away everything that isn't text with Crop image, and turn sideways photos with Rotate image. Both run in your browser.
- Have a whole scanned document? OCR PDF makes the PDF itself searchable.
Questions
Is my image uploaded?
Yes. Unlike most tools on this site, this one runs on our server. Your browser first prepares a copy of the image (turned upright and resized for reading), then sends that copy over HTTPS. The server reads the text and sends it straight back. The image exists on the server only while the job runs, in memory and in a private temporary folder that is deleted the moment the job ends. We don't store the image or the text, and we never log file contents.
How does it read the text?
With the open-source Tesseract OCR engine. Your browser scales the image so the letters are a size Tesseract reads well (the longest side between 1,600 and 3,200 pixels), our server reads it and returns the words in reading order, line by line.
Which languages work?
English, and other languages written in the same Latin letters without accents, with lower accuracy. Accented letters, other scripts and handwriting aren't recognised reliably yet.
Why are some words wrong?
OCR works best on sharp, straight, well-lit text. Blurry photos, text at an angle, decorative fonts, low contrast and text over pictures all cause mistakes. The result is editable before you copy it, so you can fix a word or two.
Can I paste a screenshot?
Yes. Copy a screenshot to the clipboard and press Ctrl+V (Cmd+V on a Mac) anywhere on this page.
Is there a limit?
One image per job, 10 free server jobs a day per person, shared by all of our server tools. Only jobs that succeed count, and the count resets at midnight UTC. For more, the OCR API is paid per call.
What is the human check?
Before a job runs, Cloudflare Turnstile checks that a person, not a bot, is using the tool. It is usually invisible; sometimes it asks you to tick a box.