Image to Text (OCR)
Last updated 22 July 2026
Drop, paste, or choose an image and pull the text out of it. Useful when you have grabbed a screenshot of a command, a repo URL, a code snippet or a slide from a video and would rather not retype it. Everything runs in your browser.
Best on clean screenshots
This is a free, in-browser tool. It reads crisp screenshots (Win+Shift+S on Windows, Cmd+Shift+4 on a Mac) near-perfectly. A photo of a screen — with glare, angle and reflections — is OCR's hard case: expect a usable rough draft you'll need to fix, not a perfect read. When the text is already on a screen, screenshot it instead of photographing it.
Your image never leaves your browser
The image is read and recognised locally on your device — never uploaded, stored, or shared. The OCR engine is served from this site itself (no third-party request), so once your browser has cached it on the first run, the tool keeps working even offline.
Image
How to use it
- Get the image in. Drag a file onto the box above, click it to browse, or take a screen grab (Win+Shift+S on Windows, Cmd+Shift+4 on a Mac) and paste with Ctrl/Cmd-V anywhere on this page.
- Wait a moment. The first run downloads the OCR engine, then reads the image; a progress line shows how it's going.
- Copy the text. The result lands in an editable box — fix any misreads, then hit Copy text.
What it's good for (and what it isn't)
This is a fast way to lift text you'd otherwise retype — a command or ticker from a YouTube video, a GitHub repo URL for a backtest, a snippet of code, a line from a slide or a document. It reads clear printed text well. It is weaker on blurry, low-contrast, skewed or handwritten images, and it can misread look-alike characters (a capital O versus a zero, l versus 1). Treat the output as a draft: read it before you rely on it, especially numbers and code.
One tip that beats any setting: if the text is on a screen, take a screenshot rather than photographing it (Win+Shift+S, or Cmd-Shift-4 on a Mac). A photo of a screen adds glare, angle and reflections — the hardest case for OCR — while a screenshot is pixel-perfect and reads near-flawlessly. The "enhance" option is there for when a photo is all you have.
Frequently asked questions
Is my image uploaded anywhere?
No. Your image is read and recognised entirely in your browser — never uploaded to our server, stored, or shared. The OCR engine (the recognition code, WebAssembly core and English model) is served from this site itself, not a third-party CDN, so no outside service is contacted at all. Your browser caches it on the first run, after which the tool even works offline.
What does the "enhance for screenshots" option do?
Before reading, it upscales small images, straightens (deskews) tilted text, and applies a local adaptive threshold — a per-region black-and-white conversion that copes with glare and uneven lighting far better than one global setting — auto-inverting dark-mode captures along the way. It helps most on photos of screens and low-contrast platform screenshots. On an already-clean, high-contrast image it occasionally reads better unticked, so try both if a result looks off.
What kind of images work best?
Clear, high-contrast printed text: screenshots of a command, a URL, a code snippet, a terminal, a document or a slide. Sharper and larger is better — if the text looks small, zoom in before you screenshot. Low-contrast, blurry, skewed, or handwritten text is where OCR struggles.
Why is some of the text wrong?
OCR reads pixels, not meaning, so it can misread similar-looking characters — a capital O for a zero, l for 1, or a comma for a period. Always eyeball the result before you paste it somewhere that matters, especially numbers, tickers and code. It is a time-saver, not a guarantee.
What image formats can I use?
PNG, JPG, WebP, GIF and BMP, up to 15 MB. You can drag a file in, click to choose one, or just paste an image straight from your clipboard after a screen grab (Win+Shift+S on Windows, Cmd+Shift+4 on a Mac).
Does it work offline?
Yes, after the first use. The engine and English model are served from this site and cached by your browser on the first run; from then on the recognition runs with no network at all. Only that very first load needs a connection.
Browse the rest of the calculators & tools, or clean a broker export with the Thinkorswim trade history cleaner.
Educational utility only — not financial advice. OCR results can contain errors; always check the extracted text before using it.