Pull text out of a picture or a scanned PDF

Extract text reads the words in a picture the way a person would, by their shape, not by a text layer that has to already be there. What it finds opens as an editable page, the same idea as a transcript, so a misread word gets fixed before anything is saved rather than after.

Try free for 14 days See pricing

What it does

Reads a photo, a screenshot, or a scanned PDF with no text layer of its own.
The result opens as text you can edit, not a file dropped straight to disk -- correct a word, then save it.
Lines the model was less sure of are listed separately, so you know what is worth a second look.
A scanned PDF keeps its page breaks, each one marked, next to the document itself so you can check a line against the original.

What you can set

Nothing to set beyond the order of a page -- open the file and press Run.

Works with

Images (PNG, JPG, WEBP, GIF, BMP, TIFF, AVIF and HEIC) and PDF.

Questions

Does it work on handwriting?

It is built and trained for printed text -- signs, documents, screenshots, receipts. Handwriting is a different, much harder problem and is not what this is for.

Why edit the text instead of just saving it?

Because no model reads every page perfectly. The recognised text opens the same way a transcript does -- correct what is wrong, then save -- rather than trusting the first pass and finding the mistake later, inside whatever it was pasted into.