Why We Built an Independent Browser Remover
For years, removing an image background required uploading your personal photos to remote cloud servers operated by third-party SaaS providers. Those providers process your images in distant datacenters, retain upload logs, impose credit paywalls, and place watermarks on your work.
Remove G was engineered from the ground up to prove that high-accuracy machine learning does not need to compromise user privacy. With modern browser standards like WebAssembly (WASM), WebGPU, and ONNX Runtime Web, the neural network executes directly inside your browser tab on your computer’s own hardware (learn more in our guide on how local browser AI isolates subjects).
Local Neural Computation
Inference runs locally via the open-source IS-Net salient object segmentation architecture. The model evaluates sub-pixel edge gradients, hair strands, and transparent glass without ever sending pixel data over the network (read our deep dive on continuous alpha matting).
Guaranteed Device Isolation
Because there is no remote processing server or SaaS backend, your confidential product shots, client photographs, and personal portraits are protected by physical architecture, not merely promises (reviewed in our Privacy Policy).
Technical Specifications & Open Standards
- Runtime Engine: ONNX Runtime Web with WebGPU and multi-threaded WASM execution providers.
- Neural Architecture: IS-Net salient object detection trained on high-resolution masks.
- Local Compositing: HTML5 Canvas 2D and OffscreenCanvas with 8-bit alpha channel transparency.
- Input & Export Formats: PNG, JPG, JPEG, and modern WebP up to 25 MB per image (see our image format export guide).
- Browser Compatibility: Works across Google Chrome, Apple Safari, Microsoft Edge, Mozilla Firefox, and Chromium-based browsers.