Remove Background Without Uploading
On your device, and you can check
To remove a background without uploading the image, the cut-out has to run on your own device. That is how this tool works: the AI model downloads to your browser and processes the photo locally, so the file never reaches a server.
Drop an image here
or browse your files, or paste with Ctrl+V
PNG · JPG · WEBP
Images are processed on your device. Nothing is uploaded.
Why use nobg.cc for this
- The image is processed in the browser tab - no server ever receives it
- Safe for unreleased products, personal photos, ID documents and client work
- No account, so there is nothing tying an image to you
- Verifiable in two minutes with your browser’s own developer tools
- Free and unlimited, because your hardware does the work
How to remove a background without uploading
- 01
Open the tool
The page and the removal library load like any other website. No image is involved yet.
- 02
Drop your image
The file is read straight into the page. The model runs on your GPU through WebGPU, or on the CPU.
- 03
Download the result
The PNG or JPG is created in the tab and saved from there. Close the tab and nothing is left behind.
How it works, and how to verify it
Server-based vs on-device background removal
Most online background removers send your image to a server, run the model there and send the result back. That is not wrong, but it means a copy of your photo leaves your device and you have to trust what happens to it. On-device removal moves the model to you instead of moving the photo to the model.
| Server-based remover | On-device (this tool) | |
|---|---|---|
| Where the image goes | Uploaded to the provider’s server | Stays in your browser tab |
| What gets downloaded | Only the result | The model, once (44-176 MB depending on the option), then cached |
| First image | Fast - limited by upload speed and queue | Slower while the model downloads |
| Every image after that | Upload and wait again | A few seconds on the GPU, no network needed for the image |
| Cost model | Server time costs money, so credits, limits or low-res previews are common | Your hardware does the work, so it can stay free and unlimited |
Check it yourself in the Network tab
You do not have to take the privacy claim on trust. Any desktop browser can show you every request the page makes:
- Open the developer tools (F12, or Cmd+Option+I on a Mac) and switch to the Network tab.
- Drop an image on the tool and watch the list while it processes.
- You will see the model and runtime files being downloaded from staticimgly.com, the CDN of IMG.LY, who publish the open-source removal library. Those are downloads: the requests go out empty and the model comes back.
- You will also see a page view and a couple of small analytics events (cookieless Plausible) that record the model and device used, the file type and a size range such as “0.5-2MB”. Never the image, never its name.
- What you will not see is any request carrying your image. Click any request and check its payload.
Who this matters for
Sellers photographing products before launch, anyone editing a photo of a person, agencies working with client assets under NDA, and anyone who simply prefers their photos to stay theirs. The privacy page has the short version, and the about page explains the model and library in more detail.
From experience
Honest trade-offs of running it locally
- The first image of a session is the slow one: the model has to download. On a slow connection the 44 MB Fast model gets you started sooner.
- Speed depends on your device. A laptop with WebGPU is fast; an older phone on CPU can take much longer on a large photo.
- Very large images use a lot of memory in the browser. If a tab struggles, resize the photo first.
- The model is cached like any other file, so clearing your browser cache means one more download next time.
Written and maintained by Piero Nanni, Developer, nobg.cc. Last updated 2026-09-26. Something wrong here? Let me know.