Only Trusted Image to Text (OCR)
Pull the printed text out of a photo, screenshot, scanned page or PDF, so you can copy it, edit it and save it as a text file. The reading happens on your device: your images are never uploaded.
Free · No uploads · On-deviceDrop images or a PDF here, or click to choose
JPG, PNG or WebP. As many as you like, read in the order you add them. You can also paste (Ctrl+V) an image straight in.
Options
English only at launch. Printed text reads well. Handwriting, blurry photos and complex layouts read poorly, and this page says so plainly rather than guessing at your text.
Engine loads only when you add a first image.
Result
Add an image and the reading shows up in the text box below.
Processed on this device: 0 · Uploaded: 0
Download
Your images
Nothing here yet. Drop images into the first tile to start.
Text found
Readings from several images are added one after another, each under a line with its file name. Always check names, numbers and totals against the original image before you rely on them.
How it works
The engine loads only when you need it
This page loads light. Nothing heavy is downloaded when you open it. The first time you add an image, your browser loads the reading engine (Tesseract, open source) and the English language data from this site, about 6 MB in total, and shows the progress above. The language data is then cached on your device in IndexedDB, and the engine files sit in your browser cache, so repeat use starts much faster. Your images stay on the device the whole time.
Guide
Getting a cleaner read
- Photograph straight on: a page shot at an angle loses words at the edges. Fill the frame and keep the page flat.
- Use good, even light: shadows across a page and glare on glossy paper both cost words.
- Prefer the original file: a screenshot reads better than a photo of a screen. Send the screenshot itself, at full size.
- Crop tight: cut out hands, desk and background before you add the image, so the engine spends its effort on the page.
- Split long pages: one clear image per page reads better than one tall photo of a whole document.
Questions
What kind of text does this tool read well?
Printed text. Clean printed pages, receipts, labels, screenshots and photos of documents in good light read well. Accuracy drops on blurry photos, low light, slanted shots and busy backgrounds, and layout is not preserved: the reading comes out as plain lines of text, in reading order, without columns, tables or font sizes.
Can it read handwriting?
Not reliably. This engine was made for printed text, and handwriting usually comes out with many wrong words or nothing useful at all. For a handwritten note, typing it out is faster and more accurate.
Are my images uploaded to a server?
No. The reading engine and the English language data load from this site into your browser the first time you add an image, and the reading itself happens on your device. Your images are never sent anywhere. After that first load, the language data is cached on your device in IndexedDB, and the engine files sit in your browser cache, so repeat use is faster.
Which images and languages does it support?
JPG, PNG and WebP images, and PDF files (up to 10 pages per PDF, read page by page), in English. Only English is included at launch, to keep the first download small. A photo of a sign or page in another language will produce garbled text, so check the wording before you rely on it.