How-to · OCR
How to extract text from an image in Chrome
The short answer
Drop the image into a local OCR workspace — recognition runs on a bundled Tesseract engine inside your browser, so the image is never uploaded. You get editable, selectable text back in seconds, ready to copy.
- 1
Click the Image OCR icon and choose "Text from an image file" to open the workspace.
- 2
Drop, paste (Ctrl+V) or pick the image — PNG, JPG, WebP or BMP.
- 3
Wait a few seconds while the bundled engine recognizes the text locally.
- 4
Check the editable result against the original and fix any misread characters.
- 5
Click "Copy all" to put the extracted text on your clipboard.
You have a photo of a document, a scanned receipt, a slide from a talk, or a meme with a quote you need — and the text is trapped in pixels. Most "image to text" sites make you upload the file to their server, wait, and hope it isn't kept. In Chrome you can skip the upload entirely: run the OCR engine locally, inside an extension, and the image never leaves your machine.
Open the workspace and drop the image in
Click the Image OCR icon and choose Text from an image file. The workspace accepts PNG, JPG, WebP and BMP — drag the file in, paste it with Ctrl+V straight from your clipboard, or pick it with the file chooser. Pasting is the underrated one: screenshot something with your OS shortcut, hit Ctrl+V in the workspace, done.
Recognition runs entirely in your browser
The moment the image lands, recognition starts. There is no cloud API behind it: the OCR engine (Tesseract, compiled to WebAssembly) and its language data ship inside the extension, so the whole pipeline runs on your device. That has three practical consequences. It works offline. It costs nothing per page. And nothing — not the image, not the recognized text — ever touches a server.
Fix, then copy
The result panel shows the original image and the recognized text side by side, and the text box is editable. OCR is very good on clean type and less good on low-contrast photos, so skim the output, fix the odd character in place, then hit Copy all. Click the original to compare when a word looks suspicious.
Switch the recognition language
Recognizing the wrong language is the number-one cause of garbage output. The recognition language is switchable right in the popup and the workspace, and the language data is bundled like everything else — switching never triggers a download, so it also works offline.
Why not an upload site?
For a screenshot of a lunch menu, an upload site is probably fine. For a contract, an ID, a medical letter or anything from work, you are sending the document to a server you don't control, with a retention policy you probably haven't read. A local OCR tool removes the question: it's free, it needs no account, and you can verify it makes zero network requests. If the text you need is on the screen rather than in a file, the same extension can also capture a region of the page and copy the text from it.