PDF OCR: extract text from scanned PDFs and images

Recognize the text in scanned PDFs and photos and turn them into a searchable PDF or a text file. Runs in your browser, free, no sign-up.

    Drop PDF, JPG, PNG files here or choose files.

    • Up to 5 files, up to 50MB each
    • Up to 20 PDF pages that need recognition per run (pages that already contain text are read directly)
    • The first run downloads language data (Korean plus English, about 4.5MB). Your files are never uploaded
    • Password-protected PDFs must be unlocked first

    A searchable PDF puts an invisible text layer over the original image, so you can search with Ctrl+F and copy text.

    Use 300 for small or faint text. Higher is slower and makes a bigger PDF.

    PDF only. Separate with commas, use a hyphen for ranges.

    Files never leave this device. Privacy policy

    Result

    Add a scanned PDF or a photo and click Convert. The text is recognized and saved as a searchable PDF and a text file. Expect a few seconds per page.

    What to do next

    PDF to Word converter

    Turn the PDF, now with text, into an editable Word file.

    Scan documents from photos

    Straighten a crooked document photo first for better recognition.

    PDF to JPG converter

    Save the pages as images.

    How it works

    Add a scanned PDF, a fax or a photo of a document, choose the language and output format and click Convert. Each page is rendered as an image and the recognition engine reads it line by line. A searchable PDF is the original image with an invisible text layer on top, so it looks the same but can be searched and copied.

    PDFs that already contain text (exported from Word, for example) skip recognition and the text is read directly. In a mixed document, only the pages without text go through recognition.

    Everything runs in your browser. The language data (Korean 1.5MB, English 2.8MB) is downloaded once and stored for later runs. Your files are never uploaded.

    Things to know

    The engine is built for printed text. Handwriting, vertical text, low-resolution scans (100 dpi or less) and busy backgrounds produce many errors. If the confidence shown on the result card is below 70%, rescan at 300 dpi or retake the photo straight on.

    Expect some typos in the output. For contracts and financial records, compare against the original. Tables are read line by line, so the cell structure is not preserved.

    Each page takes a few seconds, so up to 20 pages are recognized per run. For longer documents, run page ranges separately or split the PDF first. A searchable PDF re-embeds the page images and can be larger than the original.

    Frequently asked questions

    I want to copy text from a scanned PDF.

    Add the PDF and set the output to Searchable PDF or Text file. The text file can be pasted straight into Notepad or Word. The searchable PDF looks exactly like the original but lets you search and copy.

    Does it work on a photo of a document?

    Yes, JPG and PNG are accepted. A straight, evenly lit photo without shadows reads best. If it is crooked or dark, straighten and brighten it with Scan documents from photos first.

    The recognized text has errors.

    Try 300 dpi. Handwriting, vertical text, faint faxes and decorative fonts read poorly. For English-only documents, choose English only for better accuracy.

    Do my files get uploaded anywhere?

    No. The recognition engine (Tesseract) runs inside your browser. Only the language data is downloaded, once. Your files and results are never sent over the network.

    Related tools