Docrift

OCR PDF: read text from scans

For scans and photos, where there is no text to copy. Get editable text, then save as .txt or .docx.

Uses an AI model when you ask · file read on your device first

Loading the converter…

How to use it

  1. Choose a scanned PDF or a photo of a page.
  2. Pick up to 10 pages per run.
  3. Press Read text, then copy or download.

What leaves your device, and why

Every other tool on Docrift runs entirely in your browser. OCR is the exception: accurate text recognition needs a vision model that is too large to run in a web page. When you press Read text, each chosen page is drawn at about 1900 pixels on its long side, compressed to JPEG and sent over HTTPS to a vision-capable language model through Vercel’s AI Gateway. The text comes back and is shown here. Docrift does not store the images or the text.

So do not OCR documents you are not allowed to share with a third-party processor. For PDFs that already contain text, use PDF to Text instead; it never uploads anything.

Accuracy

Printed text in good scans comes back nearly perfect, including tables (cells separated by |) and multiple columns. Handwriting, faded carbon copies and very small print are harder. Always proofread names, amounts and dates.

Questions

Is OCR free?

Yes, within fair-use limits: up to 10 pages per run and a cap on runs per few minutes, which keeps the service free for everyone.

Which languages work?

Most major languages and scripts; clean printed Latin-script text is the most reliable. The text is returned in its original language, not translated.

Is my document kept?

No. Docrift has no database and does not log document content. Pages are processed and discarded.

Can I make a searchable PDF?

Docrift does not add an invisible text layer to scans. Download the .docx and use Word to PDF for a text-based PDF.

Related tools