OCR, optical character recognition, reads the letters in an image and turns them into real text. It saves retyping a printed letter, a receipt, a page from a book, a screenshot of a message or a scanned contract.
Which languages are supported?
English, Dutch, German, French and Spanish, each with its own trained model in Tesseract, the open-source OCR engine behind this tool. Choose the language of the text before you start. The right model knows the accented characters and common words of that language. Read Dutch or German text with the English model and letters such as ë, é and ü are misread far more often.
How accurate is it?
It depends almost entirely on the image. Clean printed text, scanned or photographed straight on in good light, is recognised very well. Accuracy drops with blur, low resolution, strong shadows, curved book pages, text at an angle, decorative fonts, text over busy backgrounds, and handwriting, which Tesseract is not built for.
For best results, scan at 300 dpi if you can. With a phone, hold it parallel to the page, fill the frame with the text and avoid shadows. Always proofread the output before relying on it, especially numbers and names.
What does the output look like?
A plain UTF-8 text file. The words and line breaks are kept in reading order; fonts, bold, columns and tables are not. Multi-column pages may come out one column after another, or with lines of the columns mixed, depending on the spacing.
Which files can I use?
Images in JPG, PNG, TIFF, BMP, GIF and WebP, and scanned PDFs. Each PDF page is rendered at 300 DPI in grayscale before it is read, and the text of all pages comes back in one file, separated by page breaks. If your PDF has selectable text already, PDF to Text is faster and exact.
Is it private?
OCR runs on our server. Your image is uploaded over an encrypted connection, processed in Germany and deleted with the text file automatically within two hours.
How big can the file be?
Files up to 50 MB are free without an account, within a daily allowance of 20 server jobs (50 with a free account); Pro takes up to 500 MB. A drawer of scanned paperwork is a job for the pipeline builder: the Scan cleanup preset compresses each PDF and extracts its text in one run. You can try it on up to 3 files for free.
How to extract text from an image
- Drop your image or scanned PDF onto the drop zone.
- Choose the language of the text.
- Click Extract text.
- Download the .txt file and proofread it.
FAQ
Can it read handwriting?
Only very neat block capitals, and not reliably. It is built for printed text.
Can I OCR a document with two languages?
Choose the main language. Words in the other language will usually still be read, with a few more errors.
Can I get a searchable PDF instead of a text file?
Not yet. The output is plain text.
Why is the output full of odd characters?
The image is probably blurred, rotated or low resolution. Straighten it with Rotate Image and try a sharper scan.