In short: Background removal separates the main subject of a photo from its background with an AI segmentation model and outputs a PNG with a transparent background.
About Remove Background from Image
Product photos for an online shop, a profile picture with a clean backdrop, or a cut-out for a slide — removing a background used to mean careful work with selection tools. This tool uses small neural networks that segment the foreground automatically: MODNet for portraits and U²-Net (u2netp) for products and objects.
Unlike most background removers, the model runs in your browser via ONNX Runtime WebAssembly. The first use downloads the model and its WebAssembly runtime from this site (about 24 MB for portraits or 16 MB for objects); after that your browser caches them, and your photos are never uploaded. Results are best with a clear subject and good contrast against the background. Hair and fine details may need softening with the edge slider.
How to use Remove Background from Image
- Choose or drop a photo.
- Pick Person for portraits or Product / object for everything else.
- Press Remove background and wait while the model runs (a few seconds on most devices).
- Adjust edge softness or choose a solid background color, then download the PNG.
Examples
Photograph a product on a table, choose Product / object and a white background to get a clean catalog-style image.
Use Person mode on a selfie and place it on a solid brand color for a consistent team page.
Frequently asked questions
Are my photos uploaded to a server?
No. The AI model is downloaded to your browser and runs locally. Your image is never transmitted.
Why is the first run slow?
The model file has to download once. Subsequent images reuse it and are much faster.
How accurate is it?
It handles clear subjects well. Complex scenes, transparent objects like glass, and fine hair against a busy background are harder for any automatic tool.
What resolution is the output?
The same as your original image. The mask is computed at a lower resolution and scaled up smoothly.
Updated November 2026 · Free, no signup · How we build and test tools