Skip to content
● Your files never leave your device. No upload. No signup. No watermark.

Extract text from an image, on your device

Pull text out of a picture. Photograph a page, screenshot a form, scan a receipt, and get selectable text you can paste anywhere. The recognition model runs on your own machine.

Drop an image here to read its text One image. Clear, straight, well-lit pictures read best.

    Live privacy check

    How to image to text (ocr)

    1. Drop in the image containing the text you want.
    2. Choose the language the text is written in.
    3. Click Extract text. The recognition model loads on first use, which takes a moment.
    4. Copy the text, or download it as a plain text file.

    Why this image tool is private

    Ordinary OCR means sending your document to somebody's server, and people run OCR on passports, contracts and medical letters. Here the recognition model is downloaded once and then runs locally, so the image itself is never transmitted.

    Other tools in this category upload your file to a server, process it in the cloud, and return a download link. That means your document sits on someone else's disk — often with a retention window measured in hours. Here, the file is read by your browser, transformed in memory, and saved straight back to your disk. You can watch it happen in the live check above: the counter for files uploaded stays at zero.

    Frequently asked

    Why is the first run slow?
    The recognition model is a few megabytes of data that your browser downloads the first time. After that it is cached, and later runs start quickly.
    How accurate is it?
    On clean, straight, high-contrast text it is usually very accurate. Handwriting, heavy skew, low light and decorative fonts reduce accuracy sharply. Expect to proofread anything important.
    Which languages are supported?
    English is included by default. Other languages load on demand, and each adds to the download because each needs its own model data.
    Does it keep the layout?
    It returns text in reading order with line breaks. Complex multi-column layouts may come out jumbled, since the tool aims for correct characters rather than exact positioning.
    Can it read a PDF directly?
    Convert the PDF to images with the PDF to JPG tool first, then run those images through this one. Both steps happen locally.