OCR Image
Turn any photo of text into copyable text. Drop in screenshots, receipts, scanned pages or signs, pick the language, and read the result — all on your own device.
Recognised text
Open a result to review and edit it before exporting.
Recognition settings
Advanced options
Summary
Recognised text
How to Read Text from an Image
Add your image
Drop a photo, screenshot, scan or PDF page into the upload area, or click to browse. A sharp, straight image gives the most accurate text.
Tip: crop out clutter and keep the text large and level — accuracy depends on the input quality.
Set the language
Choose one of the 29 supported languages, or turn on auto-detect to let the tool recognize the language for you from the list of languages.
Extract and export
Press Extract text. Tesseract.js reads the image in your browser, shows the result with a confidence score, and you can copy it or download a .txt file.
Frequently Asked Questions
Does OCR for images upload my pictures?
No. Tesseract.js runs in a worker inside your browser and reads the image locally, so your pictures are never sent to a server. Only the language model is fetched once and then cached.
How do I get the most accurate text?
Use a sharp, straight image with large text, pick the correct language, and crop out surrounding clutter. Photos taken at an angle or low resolution cost you characters.
Does it work on Arabic and Chinese text?
Yes. Arabic, Chinese Simplified, Chinese Traditional, Japanese and Korean are all available. Tesseract is strongest on printed text in a clean layout, and weaker on handwriting or heavily distorted scans.
How many languages are supported?
29 languages. Select one from the list or switch on auto-detect to have the tool identify the language from the image for you.
What is the confidence score?
A Tesseract confidence score from 0 to 100 for each text revision. A score above 90 means the words were recognized reliably; lower scores deserve a quick check.
Why do I see English words in a different language image?
Auto-detect boosts Latin text on some languages. Turn on Also read Latin words to read English words mixed into other scripts, or turn it off for pure native text.
How many images can I process?
As many as you like. Add a whole batch and each one is read with the same settings, then download every result as a .txt file or copy them individually.
What if the text is tilted or blurry?
Straighten the image before adding it, and keep it sharp. OCR reads best on clean input, so scan documents facing the right way and well lit.
Does it handle handwriting?
Tesseract is optimized for printed text. Handwriting, especially cursive, is recognized poorly, and footnotes or colored text over images need a manual read.
How much memory does OCR use?
OCR is more lightweight than the image AI models and runs in a worker, so it stays fast even on modest hardware.