Image & PDF Tools

AI Background Remover

Remove image backgrounds locally and download a transparent PNG with before/after preview. Free, fast, and built for a clean browser workflow.

JPG, PNG, or WebP up to 20 MB. Processed locally — not uploaded to ShishwaLab servers.

Upload a JPG, PNG, or WebP image to remove the background locally.

About this tool

This free background remover cuts the subject out of product shots, portraits, and simple object photos so you can place them on any backdrop.

A lightweight U²-NetP segmentation model runs entirely in your browser with ONNX Runtime (WebGL when available, WASM fallback). Your photo is not uploaded to ShishwaLab servers for processing.

Use it with the Image Compressor & Resizer when you need a smaller transparent PNG for websites, marketplaces, or presentations.

Tips for better results

  • Clear subjects on simple backgrounds usually cut cleanest; busy scenes may need a tighter crop first.
  • Very large images are capped on the long edge for memory safety — compress oversized originals if your device struggles.
  • The first run downloads the local model (~4.5 MB) and runtime files; later runs reuse the browser cache.
  • Download PNG to keep transparency. Pair with the compressor if you need a smaller file for upload limits.

How to use

  1. 1 Upload a JPG, PNG, or WebP image.
  2. 2 Click Remove Background and wait while the local model runs on your device.
  3. 3 Preview the transparent result on the checkerboard, then download the PNG or reset.

FAQ

Is this background remover free?

Yes. Remove backgrounds in your browser with no account and no paid AI API.

Are my images uploaded?

No. Segmentation runs locally in your browser. Your photo is not uploaded to ShishwaLab servers for processing.

Which formats are supported?

JPG, PNG, and WebP input. Output is a transparent PNG.

Does it work for products and portraits?

Yes. The general salient-object model handles people and many everyday objects. Complex hair or fine edges may look softer on the lite model.

Why is the first run slower?

The first visit loads the ONNX runtime and a ~4.5 MB model. After that, browser caching usually makes later runs faster.

What if my device is low on memory?

Try a smaller image, close other tabs, or compress first. The tool shows a clear message when memory or loading fails.