Tooooolbox
Nothing leaves your device

Image to Text

. Privately.

Get the text out of a photo, a screenshot or a scanned page, ready to copy and edit. Drop an image or paste it with Ctrl+V, pick the language, and the text appears in a few seconds. Recognition runs on your device: the image is never uploaded.

100% private — no upload
Unlimited
Free

Language of the text

Pick several if the document mixes languages. The first run downloads the engine and each language (4–7 MB in total), then works offline.

What OCR is good at

Optical character recognition turns the pixels of letters into editable text. It is the fastest way to reuse a quote from a book, a paragraph from a scanned letter, an error message in a screenshot, or the details of a receipt, without retyping them.

Accuracy depends far more on the image than on the engine. On a clean screenshot or a flat scan, expect almost no mistakes. On a phone photo taken at an angle, in dim light, with a curved page, errors creep in: always reread figures, names and email addresses before using them.

  1. Add the image

    Drop a photo, a screenshot or a scan, or paste an image you copied with Ctrl+V (Cmd+V on a Mac).

  2. Choose the language

    Select the language of the text, or several if the document mixes them.

  3. Copy or download

    Correct the text directly in the box if needed, then copy it or download it as a .txt file.

Photos, scans and PDFs

For a document photographed with a phone, crop it to the page first with the crop tool: the engine works better when the text fills the image. Very small text benefits from a high-resolution photo rather than a zoomed screenshot.

A PDF that already contains real text does not need OCR at all: convert it with PDF to Word to edit it. For a scanned PDF, which is only a series of images, export its pages with PDF to JPG, then drop them here.

Frequently asked questions

No. The recognition engine, Tesseract, runs inside your browser. The first run downloads the engine and the language data from this site, about 4 to 7 MB depending on the language, then everything happens on your device, even offline.

English, French and Spanish, and any mix of them for a document that contains several. Choosing the right language matters: it is what lets the engine read accents such as é, ñ or ç correctly.

Barely. The engine is trained on printed text: typed documents, books, receipts, screenshots, signs. Neat block capitals sometimes work; cursive handwriting does not.

Use a sharp, well-lit image where the text is straight and fills most of the frame. Crop out the surroundings, avoid shadows and reflections, and photograph a page from directly above. A screenshot gives near-perfect results.

Yes. Drop several images: they are read one after another and you can copy or download all the text together, in the order of the files.