Is it really free?
Yes. No signup, no daily limit, no watermark. The cost of running this is your CPU — once you've downloaded the model (~88 MB on first use) every removal after that is free and offline.
Does it work offline?
After the first image (when the model downloads and caches in IndexedDB), yes — the page works fully offline. Try it: load the tool, run an image, then disconnect your network.
Will my image be uploaded?
No. Look at your browser's network tab — the only request that leaves your machine after the page loads is the one-time model download from Hugging Face's CDN. Your image is never sent anywhere.
What if the result isn't perfect?
Try a higher-resolution source. If that doesn't fix it, swap the model under Advanced options. For very fine edges, you may want a heavier model — paste a model slug like briaai/RMBG-2.0 (note: that model is non-commercial use only).
Why is the first run slow?
The first call has to download the model (~88 MB) and warm up the WebAssembly runtime. Every subsequent image uses the cached model and is much faster — typically 1–4 seconds per image on a modern laptop.
What about batch processing?
Not yet — this tool is single-image. If you need to process many images at once, the same model can be wired up via 🤗 transformers.js in a build pipeline; ask us if that's interesting.