Image to Text reads the words in a photo, screenshot or scanned page and gives you them as text you can copy, edit and search. This is often called OCR, short for optical character recognition. It runs entirely in your browser, so the image, which is often something private like a letter or a form, never leaves your device.
How to copy text out of an image
- Drop in a photo, screenshot or scan.
- The first time, the recognition engine downloads (about 4 MB). After that it is instant.
- The text appears beside your image. Choose whether to keep the line breaks or join them into paragraphs.
- Copy it, or download it as a text file.
Getting a clean read
Recognition quality depends almost entirely on the picture, so a moment spent on the photo saves a lot of correcting afterwards:
- Fill the frame with the text. Crop away the desk and the margins first if you need to.
- Hold the camera square to the page. Text photographed at an angle reads badly.
- Use even light. A shadow across the page, or glare from a lamp, both hurt.
- Prefer a screenshot to a photo of a screen, which picks up moiré patterns and reflections.
The tool shows how confident it was, and lists individual words it doubted, so you know where to look.
What it reads well, and what it doesn't
Printed text is its strength: book pages, printed letters, signs, receipts, slides and screenshots. Handwriting is unreliable, because the engine is trained on typed characters. Very small text, faded receipts, and text printed over a busy photograph are all harder. It recognises the English alphabet, so other scripts and heavily accented words will not come out correctly.
Why this runs on your device
Most free OCR websites upload your picture to their servers. The things people actually scan are rarely trivial: passports, prescriptions, bank letters, contracts and forms. Here the recognition engine is downloaded to your browser instead, and the image is read where it already is, which is why the first use takes a moment and every use after that is private and offline.
Before and after
Cropping to just the text with the Image Cropper usually improves the result more than anything else. For a scanned PDF, turn the page into an image first withPDF to JPG. Once you have the text, theWord Counter tells you how long it is, and theMarkdown Previewer is a good place to tidy it into a document.
Frequently asked questions
Is my image uploaded to a server?
No. The text is recognised inside your browser, which matters here because the things people scan are often private: contracts, prescriptions, bank letters and ID documents.
Why does it download something the first time?
Reading text needs a recognition engine and an English language model, about 4 MB together. That is fetched once from this site, then cached, so later images are read straight away.
How do I get a better result?
Fill the frame with the text, hold the camera square to the page rather than at an angle, and use even light with no shadow across the page. Cropping to just the text before reading it helps a lot.
Does it read handwriting?
Rarely well. The engine is trained on printed text, so typed pages, signs and screenshots work well while handwriting is unreliable.
Can it read languages other than English?
Not yet. It recognises the English alphabet, which also covers most European languages roughly, but accented words and other scripts will come out wrong.
Can it read a PDF?
Convert the page to an image first with PDF to JPG, then read that. If the PDF already contains real text rather than a scan, you can simply select and copy it in any PDF viewer.
Last updated