Extract text from images, scanned documents and screenshots. Powered by Tesseract โ supports 30+ languages including Arabic, English, French, Urdu and more.
Scans, screenshots, photos of text
Choose ImageJPG, PNG, WebP, BMP ยท Max 5MB
This tool uses OCR (Optical Character Recognition) to detect and extract text from an image โ reading printed or typed text within a photo, screenshot, or scanned document and converting it into editable text you can copy, search or paste elsewhere.
A regular image file is just pixels โ there's no underlying text data, even if the image clearly shows words. OCR analyses the visual patterns in the image and recognises shapes that correspond to letters, numbers and words, effectively "reading" the image the way a person would. This is fundamentally different from extracting text from a digitally-created PDF (which already contains text data) โ OCR is specifically for situations where text exists only as an image.
OCR processing runs entirely in your browser โ your image is never uploaded to a server.
OCR accuracy depends heavily on image quality. Clear, well-lit photos with the text reasonably straight (not at a steep angle) and in focus produce the most accurate results. Low resolution, blurry, or low-contrast images (light text on a light background, for example) can lead to recognition errors. If results aren't accurate, try a higher-resolution photo, better lighting, or cropping the image with our Image Cropper to isolate just the text area, removing surrounding clutter that might confuse the recognition.
OCR algorithms work by recognising the shapes of individual characters, and these algorithms are generally trained on text that appears roughly horizontal and evenly lit. A photo taken at even a moderate angle distorts the shape of each character โ letters that should be perfectly round or straight-edged become subtly skewed, which can be enough to confuse recognition, especially for similar-looking characters (like "0" and "O", or "1" and "l"). Uneven lighting creates a similar problem from a different angle: a shadow falling across part of a page can make text in that area significantly darker or lower-contrast than text in a well-lit area, and the recognition algorithm may handle these two areas of the same image quite differently in terms of accuracy. Taking photos straight-on, with even lighting across the whole page โ even if this means taking a moment to reposition rather than snapping quickly โ tends to produce noticeably better OCR results than a quick, angled, unevenly-lit photo.
Even with good source images, OCR rarely produces perfectly error-free text, especially for longer documents โ a few misrecognised characters per page is common, particularly for unusual words, names, numbers, or specific formatting like currency symbols. Rather than expecting a perfect result, it's more realistic to treat OCR output as a strong starting point that needs a proofreading pass โ particularly checking numbers (which can be misread in ways that completely change their meaning, unlike a misread letter which is often still recognisable from context) and any text that's especially important (names, dates, reference numbers). For long documents, focusing review effort on these higher-stakes details, rather than reading every word of body text with equal scrutiny, is often a practical balance between thoroughness and time.
If you have a scanned PDF (rather than a single image) and need to extract text from it, you would first convert the relevant page to an image using PDF to JPG, then run that image through this OCR tool. For PDFs that already contain text data (not scanned), use PDF to Word instead, which extracts the existing text directly without needing OCR.
OCR accuracy depends on image quality โ blurry, low-resolution, or low-contrast images lead to more recognition errors. A clearer photo generally improves accuracy.
OCR is generally much more accurate with printed or typed text than handwriting, which varies significantly between individuals and is harder for recognition algorithms to interpret reliably.
Convert the relevant page to an image first using PDF to JPG, then run that image through this OCR tool.
The OCR engine primarily targets text in Latin-based alphabets (English and similar languages). Accuracy may vary for other scripts.
No, OCR processing runs entirely within your browser.