I needed simple local image generation without the usual setup. No virtual environments, no ComfyUI with a complex graph and installation as an exe. So i tried to push the whole thing into the browser and run it on WebGPU. It's a browser extension. You install it, then it loads model, and after that it runs on your own GPU, offline. It use text encoders, UNet, and VAE are ONNX graphs, running on the browser's WebGPU stack. Github : https://github.com/d0grr/generate-ai-images Firef...