I built I Hate Background (IHB) after noticing that most background-removal tools send every image to external servers. IHB takes the opposite path: it runs the entire segmentation pipeline locally in the browser, using a WebAssembly/WebGPU-optimized AI model. No uploads, no tracking, no logging.

Tech bits: • Browser: WebGPU + ONNX runtime for real-time inference • Frontend: ~12s average removal on mid-range hardware • Privacy: Zero data leaves the device

Thanks to Hugging Face and Transformers.js' team for support this.

Would love feedback on model performance, UX, and design.