Image to Text

No. There is no server behind this page. Recognition runs on your own computer.

The recognition engine has to download once, about four megabytes. After that your browser caches it and later images are much quicker.

Pick the language above the drop zone; the list includes Japanese, Korean, Arabic and Thai as well as Chinese and the European languages; the first time you use a less common language it takes a little longer to start; a wrong pick gives garbled words even on a sharp image. Accuracy depends on how sharp the text is; a clean screenshot reads far better than a photo taken at an angle.

Not reliably. The engine is built for printed text.

No. Add as many as your computer can handle.

No. Each run loads one language pack, the one selected in the list, and every image in that run is read with it. If you have a pile of English screenshots and a few Chinese ones, do them in two passes: English first, then switch the list to Chinese and run the rest.

Add them all before you press Start. They are read one after another, and the text comes back in a single box. With more than one image, each block of text starts with the file name of the picture it came from, so you can tell which part is which. With only one image you get the text alone, with no file name on top.

Under the result there are two buttons. Copy puts the whole text on your clipboard. Download saves it as a plain text file called text.txt. If your browser blocks clipboard access, the text in the box gets selected instead, so you can copy it with your keyboard.

The engine, its WebAssembly files and the packs for the eight most used languages are served from this site. The packs for the other languages are fetched from the public jsDelivr CDN, where tesseract.js publishes them; only the name of the language file is requested. The first run downloads them into your browser. After that the image is read on your own machine, and nothing about the picture is sent back.

Not on this page. It takes image files only. For a scanned PDF, use PDF to Word: it has its own option to read scanned pages with the same engine and the same languages, and gives you a .docx. Or turn the pages into images first with PDF to JPG and bring those here.

No. What comes out is plain text, the same kind you would type into a notepad. Bold, font sizes, colours and table borders are not part of it. Expect to tidy up spacing if the picture had several columns or a table, and check numbers and names before you rely on them.