OCR for images and scanned PDFs

Updated 28 September 2026

To pull the text out of an image, drop the file, choose the language it is written in and click Extract text. You get a .txt file with the words in reading order, ready to paste into a document, an email or a spreadsheet.

OCR (optical character recognition) spots printed letters in a picture and turns them into characters you can edit and search. It saves retyping receipts, contracts, printed letters, pages from a book and screenshots of chats.

How to extract text from a photo or scan

1. Add an image or a scanned PDF to the page.
2. Pick the language the text is written in.
3. Click Extract text.
4. Save the .txt file and read it through before relying on it.

Files and languages it reads

JPG, PNG, WebP, TIFF, BMP and GIF images all work, and so do scanned PDFs. Before reading, every page of a PDF is turned into a grayscale image at 300 DPI, and the text of all pages lands in one file, with page breaks in between. When a PDF already has text you can select, PDF to Text is quicker and exact.

Five languages are available: English, French, German, Spanish and Dutch. The OCR engine doing the reading is open source and carries a separately trained model for each, so the choice matters. A model knows the accented letters and frequent words of its own language, so Dutch or German put through the English model gets characters such as ë, é and ü wrong far more often.

Getting a clean result

The picture decides most of the outcome. Sharp printed text, captured straight on in even light, comes through very well. Accuracy falls with blur, low resolution, heavy shadows, curved pages, slanted lines, ornamental fonts and text printed over busy backgrounds.

Scan at 300 dpi if possible. When using a phone, keep the camera parallel to the page, let the text fill the frame and keep shadows off it. Then proofread, paying extra attention to names and figures.

What the text file contains

Plain UTF-8 text. Words and line breaks follow the reading order. Formatting such as fonts and bold is dropped, and so are table and column layouts. On a page with several columns, the columns may follow one another or their lines may interleave, depending on how far apart they sit.

A drawer full of scans

Drop a folder of scans and each file comes back as its own text file. Without an account a batch holds up to 5 files of up to 50 MB each, with 20 server jobs a day included. A free account allows 10 files per batch and 50 jobs a day. Pro raises the batch to 500 files, accepts files up to 500 MB and has no daily allowance. Download the results one by one or as one ZIP.

For scanned PDFs, the Scan cleanup preset in the pipeline builder compresses each file, then pulls out its text in the same run. Try it free on up to 3 files, or use credits for one large run: each credit covers 25 files and never expires.

Where the reading happens

OCR runs on our servers. Files travel over an encrypted connection, are read in Germany and are deleted, together with the text, automatically within two hours.