How to remove a background without uploading the photo anywhere
There is a category of picture you hesitate over before dragging it into a free web tool: a passport photo, a signed contract, a medical form, a photo of somebody else’s child, a product that has not been announced. The hesitation is correct. Most background removers work by uploading your file to a server. A browser-based one does not have to, and you can check the difference yourself in about ten seconds.
What "uploaded" actually means
When a site processes your image on its own hardware, your file leaves your device, crosses the network and lands on a machine you do not control. Reputable services delete it on a schedule and say so in their privacy policy. That is a promise, not a mechanism — and a promise is not much use if the picture is one you have a legal duty to protect, or one that simply is not yours to share. Under GDPR, a photograph of an identifiable person is personal data, and sending it to a third-party processor is a decision you are accountable for.
How the browser can do it instead
Modern browsers can run a neural network locally, through WebAssembly on the CPU or WebGPU on the graphics card. The model file is downloaded once — for Cutout, about 45 MB — and cached. After that the whole cut-out happens inside the tab: the file is read with the browser’s own file API, the model produces a mask, and the result is drawn to a canvas you download from. The image itself never travels.
Verifying it rather than believing it
Open your browser’s developer tools, go to the Network tab, and then use the tool. You will see the page load, and you will see the model file download on the first run. You will not see a request carrying your image, because there is not one. You can go further: load the page, turn your network off, and process an image anyway — once the model is cached it works offline, which is only possible if nothing is being sent.
What is still fetched, honestly
Being precise matters more than sounding good. The page itself is served from a web server, and the model weights come from a content delivery network, so those hosts see that a browser at your IP address asked for a static file. That is the same footprint as visiting any website. What they do not see, and cannot see, is the picture — it is never part of any request. If you want even that footprint gone, load the page once and then work offline.
Frequently asked questions
Is a browser-based background remover really private?
For the image, yes, and it is checkable rather than a policy promise: the Network tab shows no upload, and the tool keeps working with the network off once the model is cached.
Does it work offline?
After the first run, yes. The model stays in the browser cache, so a later visit can process images with no connection at all.
Is this good enough for GDPR?
Not processing personal data outside your own device removes the third-party processor from the picture entirely, which is a much simpler position than relying on someone else’s retention policy. It is not legal advice, but it is a materially different situation.
More guides
Other things we made
Also in English, made by the same people.
Also in English, made by the same people.
Also in English, made by the same people.
Also in English, made by the same people.