I am a person who does editing and digital designing. I used to use this tool, remove.bg, for getting transparent PNGs from either stock images or my own images. It was good, fast, free, and convenient. The only caveats for me were the export resolution limits and a privacy factor, that is, my images got uploaded to their servers.

Recently, remove.bg announced that they are shutting down the standalone website and moving to inside Canva only. This really made me sad. I have been using this tool for a long time now.

So, I made my own open-source tool, RMBG, which is: 1. Privacy-focused: the images never get uploaded somewhere; they stay and get processed only in your own browser. 2. No watermark or export resolution limits. 3. Fully free and open source with an MIT licence. No sign-in or accounts needed. 4. You can batch process and export as ZIP. 5. You can edit the image after the automated background removal, like erase objects, restore areas, recompute, etc. There are many view modes—mask, split, etc. 6. You can export in PNG, WebP, JPEG, and, in most browsers, AVIF. Other notes:

Behind the scenes, it uses the BiRefNet-lite AI model in ONNX through Transformers.js. Inference runs in a worker with WebGPU where available and a single-threaded WASM fallback.

The first visit downloads the runtime and about 98 MB of model weights from CDNs (that's why the first visit might be slower); later visits usually reuse the browser cache, so it can be fast.

There is minimal analytics, using only GoatCounter for small events such as page visits, successful inferences, and quality feedback. No image pixels, filenames, or dimensions are collected.

I am looking for contributors and as much help and support as possible, so I would appreciate people trying it out and giving feedback. Any ideas, PRs, and issues will be greatly appreciated.

I will be active on HN and will reply to comments here. Hope to see some support here!

Links: 1. Website - https://velmet44.github.io/rmbg/ 2. Source code on GitHub - https://github.com/Velmet44/rmbg 3. GoatCounter public analytics dashboard - https://velmet.goatcounter.com/ 4. Demo image - https://velmet44.github.io/rmbg/sample-split.png