Result will appear here.
Extract text from an image with on-device OCR — confidence and word counts, in your browser.
Recognises printed/typed text in an image and downloads it as a .txt file, with confidence and counts.
Each language loads its data from /assets/tessdata; the first use may download it, cached by your browser when allowed. If the cache is cleared or blocked it can load again.
Cropping, rotating, thresholding and other clean-up are not built in yet — crop or sharpen the image first for better accuracy.
The image is read on this page only; nothing is uploaded and the image is not modified.
Image OCR reads printed or typed text out of an image and gives you back editable plain text, using the Tesseract OCR engine running entirely in your browser. The result panel reports the OCR language, the mean confidence, and the character, word and line counts so you can judge how reliable the extraction is.
Image OCR supports English, Hindi, French, Spanish, German, Arabic, Chinese (Simplified), Japanese and Russian. Each language reads its own data file from this site; English is always available, and the other languages work only when their data file has been installed. If no readable text is found the tool says so instead of saving an empty file.
The image is read only to recognise its text — it is never modified or uploaded, and animated images are read as a single still frame. The download is a plain .txt file of the extracted text.
Text is recognised by Tesseract in your browser — the image is never uploaded.
The result reports mean confidence plus character, word and line counts, with a warning when confidence is low.
When no text is found you get a clear message instead of an empty download.
Pick from nine languages; each works when its data file is installed on the site.
Turn a screenshot of an article, error message or chat into editable text.
Extract the text from a photographed document or sign.
Pull a passage out of a picture without retyping it.
Get the source text out of an image so you can paste it into a translator.
OCR is tuned for printed and typed text. Handwriting, stylised fonts and low-contrast or skewed text are far less reliable; for the best result use a sharp, high-contrast image of printed text.
It lists English, Hindi, French, Spanish, German, Arabic, Chinese (Simplified), Japanese and Russian. Each language reads its data file from /assets/tessdata on this site; English is always available, and other languages work only when their data file is present.
No. The image is read by the OCR engine in your browser to extract text; the image itself is not modified, and no file is sent to a server.
It reads a single still frame. OCR does not run across animation frames, so the text reflects only the frame the browser decodes.
If no text is found, the tool reports it rather than saving a blank file. A blurry or low-contrast image, the wrong language, or genuinely hard text are the usual causes; the result panel shows the mean confidence so you can gauge reliability.
Tools that pair well with Image OCR.