Image to Text Converter (OCR)

Free Image to Text Converter (OCR)

Extract text from a screenshot, scan or photo without uploading anything. Recognition runs in your own browser using Tesseract compiled to WebAssembly, so the picture stays on your device — useful when the image is an invoice, a payslip, an ID or anything else you would rather not post to a stranger's server. Paste a screenshot straight from the clipboard, drag a file in, or fetch one by URL.

What it is good at, and what it is not

It does well with:

• Screenshots of typed text — documents, web pages, error messages, code, chat logs.
• Clean scans and flat photographs of printed pages.
• Dark-mode screenshots and terminal output, once inverted — which this tool detects and does for you.

It struggles with:

• Handwriting. Tesseract is trained on print, and cursive is effectively out of reach.
• Text over busy photographs, gradients or patterns rather than a plain background.
• Heavily stylised or decorative type, and very small or blurred text.
• Skewed or curved pages — straighten the image first and you will get far more out of it.
• Complicated layouts such as multi-column magazines, tables and forms, where the reading order goes astray.

Always read the output over before you use it. Digits and punctuation are where errors cluster — 0 and O, 1 and l and I, 5 and S — which matters most in exactly the places accuracy counts, like reference numbers and amounts. The words the engine was unsure about are highlighted for you, but a clean-looking result can still contain mistakes.

Getting better results

1. Start bigger. Resolution matters more than anything else. Capture the screenshot at full size rather than scaling it down first, and use the Enlarge slider for small text.
2. Crop to the text. Removing surrounding interface and decoration helps the engine find the layout.
3. Raise the contrast for faint or greyish text, and try black and white only for clean printed type. Leave it off for photographs.
4. Pick the right layout. A single line reads better as “One line” than on Automatic, and scattered interface labels read better as “Scattered text”.
5. Set the language. Accented and non-Latin text needs the matching language, and using the wrong one badly hurts accuracy.

Supported Languages and File Types

29 languages are offered here, from a Tesseract set of more than a hundred, including Chinese, Japanese, Korean, Arabic, Hebrew, Hindi, Thai and the major European languages. Images can be PNG, JPG, WebP, GIF, BMP or AVIF — anything your browser can decode.

Related tools

Need the colours out of an image instead of the words? Try the image colour picker. To tidy up a screenshot before reading it, the image cropper and sharpen tool both help.

Privacy and Security

We respect your privacy. Recognition happens inside your browser, so your image is never uploaded to our servers and nothing is stored once you close the page. The recognition engine and language data are the only things downloaded, and they come from a public software CDN — the image is never part of that request.