Copy Text from Image - Select & Copy Text Online Free

Upload your image and choose the language shown in it. The tool runs OCR, then overlays the recognized words back onto the picture as invisible but selectable text. Drag across the words to highlight them and press Ctrl/⌘+C, or click "Copy all text" to grab everything at once — exactly like selecting text on a normal web page.

A photo or screenshot is just pixels — there are no real characters underneath for your cursor to grab. This tool uses Optical Character Recognition (OCR) to detect each word and its exact position, then places real selectable text on top of every word so highlighting and copying works just like it does on a web page.

It supports 60+ languages including English, Spanish, French, German, Portuguese, Italian, Dutch, Chinese (Simplified and Traditional), Japanese, Korean, Arabic, Hebrew, Hindi, Punjabi and Russian. Pick the language that matches the text in the image from the dropdown for the most accurate reading, then re-run if needed.

Yes. App screenshots, website captures, WhatsApp chats, scanned pages, receipts, product labels and street signs all work as long as the text is readable to a human. If the result looks off on stylised fonts or tougher images, switch from Engine 1 (fast) to Engine 2 (better accuracy) and re-run.

The selectable text is positioned using the coordinates the OCR engine reports for each word, scaled to your screen. Alignment is very close but can drift slightly on heavily skewed or curved text. Copying still returns the correct words — and you can toggle "Highlight detected words" off if you prefer a cleaner view while selecting.

The tool is free with no signup and no watermarks. Your image is uploaded over HTTPS, the OCR runs on our secure servers, and the upload is automatically deleted within minutes. We never share your files with third parties or use them to train models, so it is safe for screenshots of private chats, IDs and documents.

Yes. Open the tool in your phone's browser, upload a screenshot or photo from your camera roll, and pick the language. The recognised words are laid over the image as selectable text — long-press to start a selection and drag the handles to grab the part you want, or tap 'Copy all text' to copy everything at once. No app to install.

This tool keeps the words in place on your image as selectable text, so you can highlight and copy just the part you need — like selecting text on a web page. If you instead want the whole image turned into one block of editable, exportable text, use our Image to Text converter, which runs the same OCR and returns the full recognised text ready to paste or download.

Two settings drive accuracy. First, set the source language to match the text in the image — this is the biggest factor, because the engine then expects the right alphabet. Second, if the result looks wrong, switch the OCR engine: Engine 1 is fast and great for clean screenshots, while Engine 2 is more accurate on stylised fonts, low contrast or tougher photos. Toggle 'Highlight detected words' on to see exactly which words were read and spot anything the engine missed before you copy.

Default runs PaddleOCR, a modern general-purpose recognition model, and gives the most accurate result for most photos and screenshots — it's the best starting point. Engine 1 runs Tesseract, a fast, lightweight engine that's solid on clean, high-contrast scans, while Engine 2 runs EasyOCR, which tends to do best on stylised fonts or lower-contrast images. If Default doesn't read an image well, try Engine 2 next.

Not every language is supported by every engine — for example Punjabi, Greek and Hebrew currently only work with Default or Engine 1, not Engine 2. If you switch to an engine that doesn't support your chosen language, the tool automatically falls back to English so the request still runs; switch back to Default or Engine 1 to use those languages again.

No — PaddleOCR (the Default engine) has no recognition model for Punjabi, Hebrew or Bengali, so all three currently only read correctly under Engine 1 (Tesseract). Bengali is also supported under Engine 2 (EasyOCR), but Punjabi and Hebrew are not, so if you're working in one of these languages, pick Engine 1 rather than Default or Engine 2.