Image to Text

Yes. OCR — optical character recognition — is the general name for reading printed text out of a picture, and that is exactly what happens here. The difference from most OCR websites is where it runs: the engine is downloaded into your browser rather than your file being uploaded to someone's server.

Accuracy depends on image clarity, language, layout and lighting. Check the extracted text against the original, especially names, dates and numbers.

No, and this is worth being clear about. The engine is built for printed and typed text. Handwriting recognition is a different problem needing different models, and trying it here will waste your time.

No. The text inside a table will be read, but the row and column structure is not preserved — you get the words, not a grid. The engine does not detect table structure at all, so there is no way to export usable Excel from it.

Selected file contents are processed in your browser without being sent to an application upload endpoint. Loading the site, fonts or processing libraries may still require network requests.

Open the page while online. Processing may continue in the loaded tab, but reloading, external assets, and uncached engines or models require a connection.

The language of the text in the image, not the language of this page. If you pick wrong the output will be nonsense even from a clear image. Non-English choices load English as well, so mixed documents with order numbers and email addresses still read correctly.

Yes. Each page is rendered and read in turn, and results appear page by page. Very long documents are capped at the first 100 pages, and you are told when that happens rather than being left to wonder.

You will see "No text found" rather than an error, because an image with no readable text is a normal result. Usually it means the picture is too blurry, the text is too small, or the wrong language is selected.

Because it should not have them. Recognition engines treat each Chinese character as a separate word and put spaces between all of them, which makes the result awkward to use. Those spaces are removed here — but spaces between Chinese and Latin text are kept, since those are real.

Yes. Drop in as many images as you like and they queue up, with results appearing as each finishes. If one file fails the rest continue, and you can stop the queue at any point without losing what has already been read.