Turn photos, screenshots and scanned pages into editable text. Drop, paste or choose images, pick the language, and the OCR engine reads the text right in your browser, so your images never leave your device.
Starting…
Image
Your image will appear here. Words the engine finds are outlined in green; uncertain words in orange.
Advertisement
Advertisement
How to use the Image to Text (OCR)
Add images: drop them on the box, paste a screenshot with Ctrl+V (Cmd+V on Mac), or click Choose images. You can add several at once.
Pick the language of the text. For documents that mix two languages, also choose a second language.
Extract the text: recognition starts automatically. The first run downloads the OCR engine and language data (a few MB), which your browser then caches.
Check the result: detected words are outlined on the image and uncertain ones are marked in orange. Edit the text if needed.
Copy or download the text as a .txt file. Change the language or options and press Extract text again to re-run.
How OCR turns a picture into text
OCR (optical character recognition) software looks at the pixels of an image and decides which letters they form. This tool uses Tesseract, an open-source OCR engine first developed at HP and later maintained with Google's support, compiled to WebAssembly so it runs inside your browser. Tesseract finds lines of text, splits them into words and reads each line with an LSTM neural network trained on printed text in each language. Every word gets a confidence score from 0 to 100, which this tool uses to mark uncertain words in orange.
Getting the best results
Problem
What to do
Small or blurry text
Use a higher-resolution image; capital letters should be at least 20 to 30 pixels tall. Keep Enhance image on: it upscales small images before reading.
Photo of a page
Shoot straight from above in even light, avoid shadows, and crop to the page. Tilted or curved lines lower accuracy.
Light text on a dark background
Enhance image converts to grayscale and stretches the contrast, which helps with faded scans and dark screenshots.
One line, a code or a sign
Choose Single line or Sparse text in Page layout.
Wrong or strange characters
Check that the language matches the text. Accented letters, Cyrillic or Asian scripts need the right language data.
Text from a PDF
If the PDF has selectable text, use PDF to Text, which is faster and exact. For scanned PDFs, take screenshots of the pages and add them here.
Page layout modes
The page layout setting tells Tesseract how the text is arranged (its "page segmentation mode"). Auto works for most documents and screenshots. Single column suits narrow text such as receipts. One uniform block treats the image as a single paragraph, useful for cropped quotes. Single line is for one line of text such as a title or serial number, and Sparse text finds as much scattered text as possible in any order, which is good for labels, diagrams and signs.
Frequently asked questions
Is this OCR tool free and private?
Yes. It is free with no sign-up and no limits. The OCR engine runs inside your browser using WebAssembly, so your images are never uploaded to a server. Only the engine and language files are downloaded from a public CDN the first time.
Which image formats are supported?
Any image your browser can open: JPG, PNG, WebP, GIF, BMP and, in most browsers, AVIF. For iPhone HEIC photos, convert them to JPG first. For PDFs, use the PDF to Text tool.
How can I get better OCR results?
Use a sharp, well-lit image where the text is at least 20 pixels tall, keep the page straight, and crop away unrelated areas. Turn on Enhance image for photos and low-contrast scans, choose the correct language, and use Single line mode for one-line images such as a sign or a code.
Can it read handwriting?
Only partly. Tesseract is trained for printed text. Neat block capitals often work, but cursive handwriting is usually not recognized reliably.
Why is the first run slower?
The first time, your browser downloads the OCR engine (about 3 MB) and the language data (about 2 to 15 MB depending on the language). Both are cached, so later runs start almost instantly.
Does it keep the layout and formatting?
It keeps lines and paragraphs, but not fonts, colors or tables. Turn off Keep line breaks to join wrapped lines into flowing paragraphs, which is handy when pasting into a document.
How accurate is it?
For clean printed text in a supported language, accuracy is typically very high. The confidence score shows how sure the engine is: around 90% or more usually means very few errors, while below 70% suggests checking the text carefully or improving the image.