Background Remover

Upload an image and click Remove Background — it identifies the image and removes the background automatically. Nothing ever leaves your browser.

What is the Background Remover?

A background remover takes a photo or scanned document and makes everything except the main subject transparent — the person or object in a photo, or the ink on a page, so you can drop it onto a different background, a slide, or a form without the original backdrop showing through.

Just upload an image and click Remove Background — the tool identifies what kind of image it is and picks the right engine automatically. A scanned signature or a document photographed on paper gets a fast pixel-brightness removal with no model to download; a real photo of a person or object gets separated using an on-device AI model (MediaPipe, from Google). Both run entirely in your browser; nothing is ever uploaded anywhere.

How to use it

  1. Upload one or more images — drag and drop, or click to choose files.
  2. Click "Remove Background" — the tool automatically identifies whether each image is a document/signature or a real photo, and removes the background accordingly.
  3. Use the before/after slider to check the result, then switch to Touch Up to erase or restore small areas with a brush.
  4. Choose a replacement background — Transparent, White, or any custom color — and download as PNG or JPG.

How it works, and why it's private

On upload, a quick pixel-brightness scan decides which engine an image needs: an image that's overwhelmingly plain, light paper with a little dark ink is treated as a document or signature and gets an instant threshold-based removal — pure canvas pixel math, nothing to download. Anything else is treated as a real photo and routed to MediaPipe's Image Segmenter, an open-source on-device AI model that downloads once per session (shown with a progress bar) and then runs locally via WebAssembly.

The image is never sent to Google or anyone else in either case — only the AI model file itself is fetched from a public CDN the first time it's needed. Because everything runs on your device, closing the tab discards everything and nothing you upload is ever stored or transmitted. The touch-up brush and background replacement work the same way, directly on the image data in memory.

Example

Two images uploaded together — a scanned signature and a selfie:

The signature is detected as mostly plain white paper → instant threshold removal
The selfie is detected as a photo → routed to the on-device AI model automatically
Both come out the same way: pick Transparent, White, or any color, then download

Frequently asked questions

Are my images uploaded to a server?

No. The document/signature removal is pure canvas pixel math and the photo AI model runs fully on-device via WebAssembly — your images never leave your browser, either way.

How does it know which type of image I uploaded?

It samples the image's pixels and checks how much of it is plain, light paper versus everything else — a document or signature is overwhelmingly plain background, a photo isn't. You never have to pick a mode yourself.

Why does it take a few seconds the first time on a photo?

A real photo is routed to an AI model (MediaPipe's Image Segmenter) that downloads once per session the first time it's needed. After that first download it's cached by your browser for the rest of the session.

Can I fix small mistakes after the background is removed?

Yes — switch to the Touch Up tab and use the Erase or Restore brush to manually correct any area, with an adjustable brush size.

Can I download it with a white or colored background instead of transparent?

Yes — pick Transparent, White, or any custom color before downloading. PNG downloads stay transparent unless you choose a non-transparent background; JPG always fills the background in, defaulting to white.