Remove the background from any image — right in your browser. Your photo is never uploaded to a server: the AI model runs entirely on your own device. No signup, no watermark, no file limit.

Drop an image here or click to choose

JPG, PNG or WebP · processed locally on your device

First use downloads the AI model (roughly 25 MB, one time — it is cached by your browser afterwards). On computers with a modern GPU this takes about a second per image. Works best on photos with a clear subject: people, portraits, products and pets.

Why this background remover is different

Almost every free background remover on the internet works the same way: you upload your photo to their server, they process it, and you download the result — usually shrunken, watermarked, or behind a signup wall. This tool takes the opposite approach. It runs an AI matting model (MODNet, Apache-2.0 licensed, published at AAAI 2022) directly in your browser using WebGPU and WebAssembly.

That has three practical consequences. Your image never leaves your computer, which matters if you are working with client work, product shots under NDA, ID photos, or anything private. There is no resolution cap or watermark, because there is no server cost for us to recover. And after the one-time model download, it works even on a flaky connection.

How to remove a background from an image

  1. Drop your image into the box above (or click it to pick a file). JPG, PNG and WebP all work.
  2. The first time only, wait for the AI model to download — your browser caches it afterwards.
  3. The result appears next to the original with a transparent checkerboard behind it.
  4. Click Download transparent PNG. The output keeps the full original resolution.

Tips for best results

The model is a matting network originally built for portraits, so it shines on people, pets and product shots with a clear subject — including tricky edges like hair and fur. A few things help: use the highest-resolution source you have, avoid images where the subject blends heavily into the background colour, and prefer Chrome or Edge — they support WebGPU, which makes processing many times faster than in browsers that fall back to CPU mode.

Frequently asked questions

Is my image uploaded anywhere?

No. The neural network runs inside your browser tab. You can verify this by loading the page, switching your Wi-Fi off after the model has downloaded, and processing an image — it still works.

Is it really free? What’s the catch?

It is free without limits because processing happens on your device, so it costs us nothing per image. No account, no watermark, no daily cap.

Why is the first run slow?

The model file (roughly 25 MB) downloads once and is cached. Every use after that skips the download and runs in about a second on a modern GPU.

What image formats and sizes are supported?

JPG, PNG and WebP of any resolution. Very large images (over ~4000px) may be slow on devices without a modern GPU.

Can I use the results commercially?

Yes. The underlying model (MODNet) is Apache-2.0 licensed and the output images are yours.