Remove Image Backgrounds Instantly with AI

Traditional background removal software relied on colour-keying (chroma key), where a specific colour — most commonly bright green or blue — was selected and removed. This required a controlled studio environment and failed completely on images with natural backgrounds. Even more advanced techniques like the GrabCut algorithm required manual input to define the foreground and background regions.

Modern AI background removal uses deep convolutional neural networks (CNNs) and, increasingly, Vision Transformer (ViT) architectures to perform semantic image segmentation. The AI has learned to understand what constitutes a foreground subject (a person, a product, an animal) versus background context, even in complex, cluttered scenes. This is achieved by training on datasets containing millions of images paired with hand-crafted or semi-automatically generated segmentation masks.

Our specific implementation uses a model optimised for browser-based inference via ONNX Runtime Web. ONNX (Open Neural Network Exchange) is an open format designed to represent machine learning models, enabling interoperability between different frameworks. By converting the model to ONNX format and running it in WebAssembly — and on devices with suitable graphics hardware, via WebGPU — we achieve performance that approaches native application speed, all within your browser tab.

One of the most technically challenging aspects of background removal is handling alpha matting — the smooth, semi-transparent transition between foreground and background. Human hair is the canonical example: individual strands, each potentially a few pixels wide, blending at their edges with the background colour. Our model is specifically trained to handle these trimap-free alpha matting scenarios, producing natural-looking edges without the halo artefacts common in older tools.