Background Remover

Erase the background of any photo in your browser. Returns a transparent PNG. First run downloads a ~80 MB AI model — subsequent runs are instant.

Files stay in your browser. Nothing is uploaded.

About this tool

Cut the subject out of any photo and download it on a transparent background — no server, no upload, no signup. The first run downloads an AI segmentation model (~80 MB) and caches it in your browser, so the second photo onward processes in 2-5 seconds. Works best on photos with a clear subject (one person, a product, a pet) on a reasonably distinct background. Complex backgrounds, fine hair, and translucent objects (glass, smoke, mesh) still trip up every browser-side model — for those a paid service or manual masking in Photoshop will give cleaner edges. Replace the background with solid white or chroma green right inside the tool if you need a quick studio shot or compositing source.

Frequently asked questions

Why does the first run take so long?
A neural network and its weights have to download (~80 MB). After the first run your browser caches them, so subsequent runs are seconds. Refreshing the page does not re-download; clearing site data does.
Does my photo get uploaded?
No. The model runs entirely in your browser tab via WebAssembly. The photo never leaves your device. Confirm with DevTools → Network: you will see the model files load once, then nothing photo-related goes out.
The edges are jagged or missing fine hair — why?
Browser-side segmentation models are smaller and faster than the cloud ones (Remove.bg, Photoshop). They handle 80-90% of photos well but struggle with fine hair, fur, transparent objects, and busy backgrounds. For perfect cutouts on portraits, a paid service still wins.
Will this work on a phone?
Yes on modern phones (iPhone 11+, recent Android flagships) but slowly — model download is ~80 MB and inference takes 10-30s on slower hardware. Cellular data charges may apply for the first run; on Wi-Fi after that, no extra data.
Can I batch multiple images?
v1 is one image at a time. The model stays loaded between runs in the same tab, so successive single-image runs are fast — just drop the next image and click process again.
What output sizes / aspect ratios are supported?
Any reasonable size. Very large images (>4000 × 4000) take longer and use more memory; the model itself works at lower resolution internally and upscales the mask, so detail at the edges depends mostly on the model, not your input size.
How does this compare to Remove.bg?
Remove.bg uses a larger cloud model and typically produces cleaner edges, especially around hair. The trade-off: their free tier has a low-resolution cap and counts each cutout. This tool has no cap, no upload, no account — but accepts slightly less perfect edges as the trade.

Related tools