Image to Text (OCR)
Extract text from any image — screenshots, scans, or photos — with an AI model that runs entirely in your browser. Free, unlimited, and nothing is uploaded.
Drop an image here, or click to browse
Processed locally — nothing is uploaded
How to extract text from an image
Drag a photo, screenshot, or scanned document into the box. Clear, well-lit pictures of printed text work best. Your image stays in this tab and is never uploaded.
Pick the text language — the page language is pre-selected, and you can switch to any of 14 supported languages. The first time, the AI model (~4 MB) downloads once and runs entirely in your browser.
Click Extract text. The recognized text appears below the image with each word highlighted. Copy it to the clipboard or download it as a .txt file.
Know-how
From paper archive to searchable text: why OCR beats retyping
Every scanned document is a dead end for search until its text is extracted. A folder of scans of contracts, receipts, or meeting notes can only be searched after OCR turns pixels back into characters. This tool does that in the browser: drag in the scan, copy the text, and paste it into your notes or document system, where it becomes findable by any keyword. Because recognition happens on your device, private paperwork never touches a third-party server — a real concern for invoices, tax documents, and anything under an NDA. The workflow that used to mean retyping pages now takes seconds, and the word highlights on the image let you verify the reading before you trust it. It is essentially an image to text converter that lives inside your tab — no upload, no queue, no page reloads.
Screenshots are the new copy-paste: quotes and notes from anywhere
Apps, PDF readers, and websites often block text selection. The workaround has always been retyping — until you realize a screenshot can be converted in seconds. Capture the paragraph, drop it here, and the text lands in the output box ready to paste into a document, chat, or translation tool. Researchers use this daily to pull quotes from papers and locked PDFs; support agents use it to grab error messages and order numbers from screenshots. The word-box highlight overlay shows exactly what the model read, so you can spot a misread character before copying it into something important.
What OCR can and cannot read: set expectations before you scan
The model reads printed text in good conditions: enough contrast between text and background, a roughly straight angle, and letters at least 10-15 pixels tall. Photos taken at a sharp angle, motion blur, decorative fonts, or text over busy backgrounds will produce errors — the confidence score reflects that honestly. For the best result, crop the image to the text itself and make sure it is in focus. And remember the honest limit: this model does not read handwriting. If you need cursive notes transcribed, that is a different, harder problem; this tool is optimized for the printed word.