Remove the background from a photo

The AI model is downloaded once (from Hugging Face) and then runs on your own device. Your photograph is never sent anywhere, which is unusual for this kind of tool.

Loading the tool…

How it works

  1. Drop a photo and pick a model: the general one for any subject (BiRefNet), or a lighter, faster one for people (MODNet).
  2. Choose a transparent or solid-colour background and how much to soften the edge.
  3. Run it and download the PNG.

Why nothing is uploaded

Every operation on this page is done by code running inside your browser tab, using the same engine that renders web pages. The file is read from disk into your tab’s memory, transformed there, and written back out as a download. It is never sent anywhere — not to us, not to a third party.

Verify it yourself

  1. Open your browser’s developer tools (F12) and select the Network tab.
  2. Load your file and run the tool.
  3. The only requests you will see fetch the tool’s own code — and, for a few heavy tools, their open-source engine from a public CDN — plus one small page-view ping to loreatec.jp (page address and title, nothing more). None of them carry your file.

Proof it stays local →

Frequently asked questions

How good is it?

Good on a clear subject against a distinct background. Weaker on fine hair, glass, chain-link and motion blur — that is a limit of automatic cut-out tools in general, not of this one. The result is shown next to the original so you can judge before using it.

Why do other sites do this better?

Because they run much larger models on their servers, and the price is that your photograph goes to them. The models here are 4–25 MB because they have to be downloaded to your browser and run there. It is a real trade-off: a smaller model recovers less detail.

Which model should I choose?

Start with the general one: it handles products, objects and animals. Switch to the portrait model for people — it is trained on people and does noticeably better on shoulders and hair.

Which models are these, and under what licence?

Two open models under permissive licences (MIT and Apache-2.0), loaded from a public model hub — the exact names are on the licences page. We deliberately avoided the popular alternatives: the best-known browser library for this is AGPL-3.0, and the best-known model may not be used commercially.