Is online image editing safe? What actually happens to your file

Two architectures, one identical-looking page

A browser-side tool loads JavaScript, decodes your file with APIs the browser already has, and writes the result back to a download link. Your image never crosses the network.

A server-side tool sends your file to a machine somewhere, processes it there, and sends a result back. This is not automatically bad — it is the only way to do some jobs — but it means a copy of your image existed on someone else's disk, governed by a retention policy you have probably not read.

From the outside the two look the same. Both show a drop zone and a download button.

How to check in thirty seconds

Open your browser's developer tools and select the Network tab before you add your file. Then run the tool and watch.

• A request appearing with a size close to your image's size means the file was uploaded.
• No new request at all, or only tiny analytics pings, means processing was local.
• A harsher test: disconnect from the internet after the page has loaded. A genuinely local tool keeps working.

Questions worth asking of any tool that does upload

If a tool has to upload — OCR, some format conversions, model-based background removal — the policy details are what matter.

• How long are files retained, and is deletion automatic or on request?
• Are uploaded files used to train anything?
• Is the download link guessable, or tied to your session?
• Is there a named operator and a working contact route for deletion requests?

What metadata gives away, whichever route you take

A photo straight from a phone can carry GPS coordinates, the exact capture time, and the device serial. Sharing it publicly shares those too. Most editing operations preserve EXIF unless the tool explicitly strips it — so if a picture is going online, check the metadata, not just the picture.

How this site is built

The image tools here decode and re-encode your file using the browser's canvas and file APIs on your own device. There is no upload endpoint for image data, so there is nothing for us to retain. Tools that cannot be built this way — the PDF and background-removal work — are labelled as unavailable rather than quietly routed through a server.