Image to Text (OCR)

Beta

Recognize and copy text from an image using OCR.

or drop PDF here

Processed securely on our server, then auto-deleted within the hour.

Read the text out of a photo of a document, on your device. Image to Text runs on-page optical recognition over a screenshot, a scan or a picture of a page and gives you the words to copy. It never uploads the image — the engine is downloaded on first use and then works locally, which is why it takes a moment the first time. Clear, high-contrast pictures read well; a blurred phone photo in poor light does not.

How to use Image to Text (OCR)

Click "Select image files" and choose the picture.
Click "Extract Text". The first run downloads the OCR engine, so give it a moment.
Copy the recognised text, and check names, numbers and codes against the image before relying on them.

Who it is for

Image to Text is for words trapped in a picture: a screenshot of a chat, a photographed receipt, a slide captured on a phone, a scan that arrived as an image rather than a PDF. It is the quick way to get those characters into something you can edit without retyping them.

How it works

The recognition engine is downloaded on first use and then runs locally, so the image itself is never transmitted.
The picture is read as a whole and the recognised characters are returned as text you can copy.
Clear, high-contrast pictures read far better than photographs, because recognition is matching shapes rather than understanding meaning.

What it cannot do

Layout is not preserved — columns can run together and tables lose their structure.
Handwriting, decorative type, skew, shadow and faint print all reduce accuracy, with proper nouns and numbers the usual casualties.
The engine is loaded for English. Other Latin-script languages often work; other scripts generally do not.
For a PDF, OCR PDF is the better tool: it makes the document itself searchable rather than giving you a flat text dump.

Frequently asked questions

Can I use it on a PDF?

Use OCR PDF for that. It adds an invisible text layer to the PDF so the document itself becomes searchable, which is more useful than copying text out of a picture of one.

Why is some of the text wrong?

Recognition reads shapes, not meaning. Clean printed text is usually reliable; handwriting, decorative typefaces, skew, faint print and shadow all reduce accuracy. Proper nouns and numbers are the usual casualties, so verify anything you will act on.

Does it keep the layout?

No — you get the words, not the page. Columns may run together and tables lose their structure. If you need the layout, keep the image; if you need the data in a spreadsheet, retype it from the image rather than trusting a flat text dump.

Which languages does it read?

The engine is loaded for English. Latin-script European languages often work reasonably; scripts that need different training data, such as Arabic or Chinese, will not.

Is my image uploaded?

No. Recognition runs in your browser, so a photo of an ID, a letter or a medical note never leaves your device.