Image to Text OCR: Read the Words of a Photo or Scan

How to use it

Choose a photo or scan of the text with Choose a file, or paste one with Ctrl+V.
Pick the Language of the text: one of the eleven languages, or a pair such as English and German for a mixed page.
Press Read the text and wait. The bar shows the steps; Cancel stops the reading at any time.
Check the words against the picture, then press Copy, or Download to keep a .txt file.

How the text is read

The reader is Tesseract, the open-source OCR engine, compiled to WebAssembly and run by tesseract.js inside a web worker, so the page stays responsive. The engine first finds the lines of text in the picture, then reads each character with a neural network trained on printed text in the chosen language. The language files are the fast models of tessdata_fast: a little less exact than the largest models, and much smaller, so the first reading starts sooner.

What makes a reading good

Large, sharp, dark text on a plain background reads best. A photo taken at a slant, in dim light, with a shadow across the page, or with very small print reads worse. The picture is read at no more than 2400 pixels on its longer side, which is enough for printed text. A very small print needs a closer photo rather than a larger file.

What this tool does not read

The reader is made for printed text. Handwriting and decorative fonts are not read reliably. A language that is not on the list, such as Korean, Persian, Urdu or Traditional Chinese, has no reader here, and Chinese is read in its Simplified form only. A PDF is not a picture: a PDF that already holds text can be read with PDF to Text. The confidence figure is the reader's own average estimate for its words. It is a guide, not a check.

Example: a page in both languages

Choose English and Turkish in Language of the text, then press Read the text. The words of both languages come out in one box, line by line in the order they lie on the page.

A low confidence figure is shown with a warning. Then compare the box with the picture, line by line, before you copy it.

Frequently asked questions

Why are some letters wrong in my text?

The reader works from the shapes of the letters, so a small, blurred, tilted or shadowed word can come out with a letter changed, for example a capital I for a small l. Check the words against the picture, and take the next photo straight on and sharp.

Why do I have to choose the language?

The reader uses a model for each language. The Turkish model reads the letters ç, ğ, ı, ş and ü, the German model the umlauts ä, ö, ü and ß, the Spanish model ñ and the accented vowels á, é, í, ó and ú, the French model é, è, ê and ç, and the Portuguese model ã, õ and ç. An English reading may not get these letters right. Choose a pair such as English and German for a page that mixes them.

Can it read a PDF or a handwritten note?

Not directly. A PDF that already holds text can be read with PDF to Text. A scanned PDF has no text, so read its pages with PDF OCR instead. Handwriting is not read reliably, because the reader is made for printed text.

Is the text kept private?

Yes. The picture is read by a worker in your browser, and neither the picture nor the text is sent to this site or to anyone else. The engine's files come from this site and are kept by your browser's normal cache; they hold no picture and no text.

Why does a reading take a while?

The first reading downloads the engine and the language files, several megabytes in all, and a large picture takes a few seconds to read. Later readings are quicker because the files are already in your browser. A reading that runs past two minutes is stopped, and you can cancel it at any time.

Why is the confidence not 100 percent?

The confidence is the reader's average estimate for the words it read. A high figure is a good sign and a low one means the text needs a check. It does not prove that every word is right.

Sources

Figures and text last checked on Wednesday, November 11, 2026.