Existing background removal tools either charge per image, require uploading photos to third-party servers, or demand account sign-ups. For users who value privacy or need quick edits without friction, these trade-offs are unnecessary barriers.
ClearCut runs a full AI inference pipeline directly in the browser using WebGPU acceleration (with WASM fallback). The BRIA RMBG-1.4 model is downloaded once, cached locally in IndexedDB, and all processing happens on-device. No server, no uploads, no cost.
Images never leave your device - zero server transmission
WebGPU-accelerated processing with instant repeat visits
No per-image fees, no sign-ups, no watermarks
ML inference runs in a dedicated Web Worker thread, keeping the UI completely responsive during processing. Transferable objects enable zero-copy data passing.
Automatically detects WebGPU support for GPU-accelerated inference, falling back to WASM+SIMD on older browsers for universal compatibility.
The 44MB ONNX model is cached in IndexedDB after first download. Repeat visits load the model in under 500ms with zero network requests.
The AI generates a probability mask that gets scaled and applied to the original image via Canvas API, producing a clean transparent PNG.
Built with Next.js 16 and React 19, statically exported for zero-server deployment. Tailwind CSS v4 handles the responsive UI with smooth animations and a before/after comparison slider.
Transformers.js provides the high-level ML API while ONNX Runtime Web handles the actual inference. The BRIA RMBG-1.4 model is quantized to int8 for efficient browser execution across WebGPU and WASM backends.
Privacy-first AI background removal - no uploads, no fees, no compromises