Type "remove background from image" into Google and you'll get a wall of near-identical tools: remove.bg, Pixlr, Picsart, Slazzer, Adobe Express. They all work the same way under the hood — you upload a photo, it travels to a server, an AI model on that server cuts out the subject, and a result comes back down. That round trip is invisible to most people, but it's worth understanding before you upload something personal.
Why almost every tool works this way
Background removal used to require serious GPU compute — the kind of AI model that doesn't fit in a browser tab. Running it meant renting server infrastructure, which is why every mainstream tool defaults to a cloud upload-and-process pipeline. It's not malicious, it's just how the tooling was built for the last several years.
What's actually changed
Modern browsers can now run real machine-learning models locally through WebAssembly and WebGPU-backed runtimes like ONNX Runtime Web. That means the same kind of subject-detection model that used to require a server can now run entirely on your own device — no upload, no server-side processing step, no photo ever leaving your browser tab. ToolKloud's background remover uses exactly this approach.
How to tell which type of tool you're using
It's usually not obvious from the homepage. A few tells: if the tool asks you to create an account or shows a processing/upload progress bar tied to a network request, it's server-side. If it works with your device offline after the page loads, or the browser's network tab shows no outgoing request containing your image data, it's processing locally. When in doubt, open your browser's DevTools Network tab before uploading — you'll see immediately whether your photo leaves your machine.
Does it matter for casual use?
For a product photo you're about to post publicly anyway, probably not much. It starts to matter more for anything you wouldn't want stored on a third party's server indefinitely — ID photos, personal portraits, images with visible location or personal details in the background, or anything under an NDA. Most server-side tools' privacy policies allow temporary storage for processing and "service improvement," which is broader than most users realize when they click upload.
The trade-off
Local, in-browser processing isn't free of downsides — it depends on your device's hardware, can be slower on older machines, and the model itself is more constrained in size than what a beefy cloud server could run. In practice, for typical photos, the quality gap has narrowed a lot. If you want the fastest possible cutout regardless of privacy, a cloud tool may still edge out on complex edge cases (hair, fur, glass). If you'd rather your photo never leave your device, a local tool is the only real option.