How to remove a background without uploading your image

What no upload means

A background remover needs the pixels of your image to estimate which parts belong to the subject. A hosted service sends those pixels to a server. A local browser remover keeps that inference step on your device.

With bgcut.dev, the browser downloads the application, ONNX Runtime files, and a model. Your source image, decoded pixels, predicted mask, and exported result stay on your machine during background removal.

Local processing does not mean the page makes no network requests. The site still has to load its own code and model files. The privacy distinction is about whether your image becomes part of a request to a background-removal server.

Remove a background locally

Open bgcut.dev, choose or paste a JPEG, PNG, WebP, or AVIF image, and wait for the result. The browser tries WebGPU first when it can use it. If that path cannot run, bgcut can retry with WebAssembly.

After inference, bgcut restores the matte to the source dimensions and produces the transparent result in the browser. You can inspect the original and result before copying or downloading the PNG.

There is no account step, API key, or per-image request to a bgcut inference service.

How to verify the claim

If privacy matters for a specific image, verify the behavior instead of relying on a badge. Open the browser developer tools, switch to the Network panel, clear the log, then process a disposable test image.

You should see application, runtime, and model requests. You should not see the test image sent as a request body to bgcut for inference. A local tool should also keep working for another image after the required runtime and model files are cached, subject to normal browser cache behavior.

For sensitive work, use your organization's normal browser and device controls as well. Local inference reduces one data transfer. It does not replace endpoint security or access policy.

When browser-local removal fits

Browser-local removal is useful when you want an interactive tool and do not want to install a package. It is also useful when the source image should not be sent to an inference provider.

For automated jobs, folders, or application code, use bgcut's CLI or Node.js API instead. Those interfaces also run locally and avoid turning browser automation into a batch-processing system.

• Use the browser for interactive one-image work and visual inspection.
• Use the CLI for shell scripts, files, and directories.
• Use the Node.js API when background removal is part of an application.