How to use Image to text (OCR)
- 1
Add images
Drop, browse or paste (Ctrl+V) one or more screenshots, photos or scans.
- 2
Set the language
English is preselected; add up to four languages when an image mixes them.
- 3
Review the text
Edit the result in place and check doubtful words in the Word boxes view.
- 4
Copy or save
Copy the text, or download a .txt file for one image or for the whole batch.
Limits and Pro
Up to 10 images per batch, each up to 25 MB, in any of the 123 languages, with up to four languages at once.
Batches of up to 500 images and files larger than 25 MB; recognition quality is identical on both plans. See Pro
About Image to text (OCR)
Image to text reads printed words in a picture and gives you plain, editable text. The output is UTF-8 text with line breaks where the OCR engine found them, which you can correct in place, copy, or save as a .txt file — one per image, or one combined file with a heading for each picture. A word-box view outlines every recognised word on the original and colours it by confidence, the fastest way to proofread a long result.
It suits screenshots of chats and error messages, scanned letters, receipts, slides, book pages, street signs and memes, in 123 languages and scripts. Before recognition the image can be enlarged when the type is small, converted to grayscale, reduced to pure black and white for photos of paper, and inverted automatically when light text sits on a dark interface. Five page-segmentation modes cover full pages, single columns, one-line crops and scattered text.
Tesseract's LSTM engine is built for printed type, so cursive handwriting, heavily stylised logos, curved lettering and very low-resolution photos give poor results. Tables come out as lines of text rather than cells, and multi-column layouts are approximated, not reproduced. For scanned PDF documents, the OCR PDF tool keeps the page structure instead.
Privacy
Recognition happens in a tesseract.js Web Worker inside your browser, so your file never leaves your device — useful for screenshots containing messages, addresses or account numbers. The OCR engine and each language's data (1–15 MB) are fetched from the jsDelivr CDN the first time and cached; no image or recognised text is sent anywhere.
Features
Paste screenshots
Ctrl+V anywhere on the page reads an image straight from the clipboard.
Preprocessing
Upscaling, grayscale, black & white and automatic dark-mode inversion before reading.
Word confidence
Every word outlined on the image and colour-coded by how sure the engine is.