OCR (Image to Text)
Read text out of a photo or screenshot. Nothing you choose ever leaves this tab.
Drop images here
JPG, PNG, WebP, or HEIC. Add as many as you like — they stay on this device.
Your file stays on this device. Nothing is sent anywhere.
What this tool does
OCR (Image to Text) reads the text out of a photo or screenshot and gives it back to you as plain, selectable text — the kind you can paste into an email, a document, or a search box instead of retyping by hand. It's built for the everyday version of this problem: a photo of a printed page, a screenshot of an error message, a picture of a whiteboard or a sign.
You can run a whole batch of images in one go, and each one gets its own text result — read on screen, copied with one click, or downloaded as a plain .txt file.
How to use it
Choose one or more images — JPG, PNG, WebP, or HEIC. Press "Read text", and each image processes in your browser tab with a progress bar tracking the batch.
Once a result is ready, the recognized text appears right there on the page — press "Copy text" to grab it straight to your clipboard, or use the download button to save it as a .txt file alongside the other formats every tool here offers.
Why nothing leaves your device
Reading text out of an image is real, meaningful computation — recognizing shapes as letters and words — and it all runs using a language model loaded once into your browser tab, not sent to a server for processing. Open your browser's network tab while running OCR and you'll see it load that model from this site the first time (a one-time, several-megabyte load, same as the HEIC decoder some other tools use), and nothing else — no request ever carries your image or the text read out of it.
Once loaded, every image you run after that is read entirely on your device. Nothing about the images or the text is kept anywhere once you close the tab.
What OCR handles well — and what it doesn't
This tool does best with clear, upright, high-contrast text — a scanned or photographed document, a clean screenshot, a sign photographed straight-on. Printed text in a common font, at a reasonable resolution, is where OCR is most reliable.
It does worse with handwriting (which general-purpose OCR is not designed to read reliably), text that's blurry, low-resolution, or photographed at a steep angle, and unusual or highly stylized fonts. If a result looks off, try a clearer or higher-resolution photo of the same text before assuming the tool got it wrong — image quality going in is the single biggest factor in how accurate the text coming out will be.