Text Recognition (OCR)
No more retyping: text recognition for photos and scanned images, German and English. It runs on your device and the picture stays there.
Checking access …
Recognition and the language data run on your device. The image is not uploaded and no third-party site is contacted.
Guide: Text Recognition (OCR)
A photographed receipt is an image to a computer and nothing else. The amount on it cannot be searched, the address cannot be copied, the invoice number cannot be carried into a form — even though all of it is sitting there perfectly legible. Text recognition closes that gap: it reads the letters out of the picture and hands them back as text.
This tool recognises German and English text and does the whole job on your device. Even the recognition data is served by this website rather than by a third party — so opening the tool opens no outbound connection, and the picture itself is not transmitted in any case.
How good the result is depends almost entirely on the input. A straight, sharp, well-contrasted shot yields text you only have to skim. A tilted photo under artificial light yields fragments. It is therefore worth more to retake the picture than to patch up the output.
Typical use cases
Carrying over invoice details
Read the invoice number, the amount and the tax rate out of a receipt instead of retyping them — with the usual transposed digits nobody notices.
Capturing business cards
After a trade fair twenty cards sit on the desk. Photographed and recognised, the contact details are in the address book within minutes.
Making old papers searchable
A file of documents is only really digitised once its contents can be searched — not merely once a picture of it exists.
Quoting from printed sources
A paragraph from a brochure, a standard or an official letter lands in the quotation as text, without transcribing it word by word.
Rating plates and serial numbers
In a photo of a machine's rating plate the number is legible but not copyable. Recognising a small crop is exactly what this is for.
How to use it
Choose the image
Pick a photo, a screenshot or a scanned image. PDFs are not read here — in that case convert the page in question to an image first, ideally at 300 dpi.
Set the language
Choose the language the text is written in. Both at once is possible and sensible for mixed material, but costs a little accuracy and time — for purely German text, stay with German.
Start the recognition
On the first run the engine is loaded, which takes a moment depending on your connection; after that the button shows the progress as a percentage. For a single image the recognition itself usually takes a few seconds.
Review and take the result
The recognised text appears in an editable field — numbers and proper names are worth a quick read, because that is where confusions are likeliest. Then copy it or save it as a text file.
What happens to your data
Recognition runs entirely on your device: the image is not uploaded, and the software and language data needed for it are served by this website rather than by a third party. Opening the tool therefore contacts no other site — which counts most when the material shows customer details or amounts.
Frequently asked questions
Why is the recognition wrong in places?
Usually it is the input: blur, shadows, a tilted shot or a patterned surface cost more accuracy than any setting can win back. A second photo, straight and filling the frame, almost always beats reworking the text.
Can it recognise handwriting?
No. The recognition is built for printed type; with handwriting the results are not usable. That is true of practically every tool of this kind that works without a server.
Why does the first start take longer?
On the first run the browser loads the recognition software and the language data. That is a few megabytes, cached afterwards — the second run begins immediately.
Is the layout of the original kept?
Only roughly. Line breaks usually survive, while columns and tables become running text. For tabular material, plan for some rework.