How to check that your images are never uploaded

Every page on this site says your photos are processed on your device and never uploaded. That is a strong claim, and you should not have to take it on trust. Your browser has built-in tools that show every request a page makes. This guide shows how to use them to check, in about five minutes, that no image data leaves your computer.

What you should expect to see

When you use the tool, the browser legitimately makes some network requests. Knowing them in advance makes the check easy:

• The page itself: HTML, JavaScript, CSS, fonts, and icons from this website.
• The AI model and its engine, the first time: a set of files including a large .onnx file (about 28MB for the default model) and ONNX Runtime's .wasm and .mjs files. Depending on the site's configuration, these come from this website, our file storage at models.imgcutout.com, huggingface.co, or cdn.jsdelivr.net. These are downloads (GET requests): data comes to your device, not from it.
• Portrait retouch models, if you use retouching: MediaPipe .task and .tflite files and their .wasm runtime, downloaded from this website.
• Advertising and consent scripts from Google domains, such as pagead2.googlesyndication.com and fundingchoicesmessages.google.com, on pages that show ads. These load ads and the consent message; the tool never passes your image to them.

What you should not see: any request that sends your image to a server. That would typically be a POST or PUT request with a large request body (roughly the size of your image file or larger), or any request containing image data.

Check in Chrome or Edge

1. Open the background remover in a new tab.
2. Open DevTools: press F12, or Ctrl+Shift+I on Windows and Linux, or Cmd+Option+I on Mac. Select the Network tab.
3. Tick Preserve log so requests are not cleared, and click the clear button (the circle with a line through it) to start with an empty list.
4. Drop a photo into the tool and wait for the result. Download the result as well.
5. Look at the list. Add the Method column if it is not shown (right-click the column headers and choose Method). You should see GET requests only, plus possibly the model files the first time.
6. To be thorough, type method:POST in the filter box, then method:PUT. With a clean list, you should see no requests from the tool. (Ads may make small POST requests to Google's ad servers for measurement; click any of them, open the Payload tab, and you will see short form fields, not image data.)
7. Sort by the Size column. The largest entries should be the model downloads (on the first run) and nothing that matches your image upload.

Check in Firefox

1. Open the tool, then press F12 (or Cmd+Option+E on Mac for the network panel directly) and select Network.
2. Enable Persist Logs in the settings gear menu and clear the list.
3. Process a photo and download the result.
4. Use the filter box with method:POST and method:PUT, or sort by the Method column. Click any request and look at its Request tab to see what was sent.

Check in Safari

1. In Safari settings, under Advanced, enable "Show features for web developers" (or "Show Develop menu" in older versions).
2. Open the tool, then choose Develop, Show Web Inspector, and select Network.
3. Process a photo and inspect the list in the same way.

The offline test

An even simpler check: if the processing happens on your device, it should work without a network connection.

1. Open the background remover and process one photo, so the page and the AI model are loaded and cached.
2. Disconnect from the internet: turn on airplane mode or turn off Wi-Fi. (In DevTools, you can instead set the Network throttling dropdown to Offline.)
3. Without reloading the page, drop in another photo.

The new photo is processed and you can download the result, with no connection at all. A tool that uploads images to a server cannot do this.

There are honest limits to the offline test: if you reload the page while offline it will not load, because the site does not install an offline app; switching to a different model quality in Settings needs that model to be downloaded; and portrait retouching needs its own small models to be downloaded once, so use it once while online if you want to test it offline. On phones and other low-memory devices, the tool also unloads the AI engine between images to save memory and reloads it from the browser cache for the next one, so offline processing there depends on the browser having kept those files. The contact form, of course, needs a connection.

Why we built it this way

Background removal often involves personal photos: faces, children, homes, documents, unreleased products. Processing them on your device means there is no copy of your image on anyone's server, no retention policy to trust, no possibility of a breach exposing it, and no use of your photos for training someone else's model. It also means you can verify the claim yourself, which is the point of this guide. More detail on how the processing works is in how on-device background removal works, and our Privacy Policy lists everything the site and its providers do receive.

If you see something unexpected

If you find a request that looks like it sends image data, please contact us with a screenshot of the request (its URL, method, and size). We take this seriously and will investigate and publish the result.

Summary

Open your browser's Network panel, clear it, process a photo, and look for POST or PUT requests with large bodies: there are none from the tool. The only large transfers are the one-time downloads of the AI model to your device. For an even simpler test, process a photo in airplane mode after the page has loaded.