lxfater/inpaint-web

inpaint-web is a browser-based tool for removing unwanted parts of a photo and for increasing image resolution. Inpainting refers to erasing an object or area from a photo and having the software fill in what the background likely looked like, so the object appears to have never been there. The tool also supports super-resolution, which takes a lower-quality image and produces a sharper, more detailed version at a larger size. What makes this project different from many similar tools is that everything runs inside your web browser with no server involved. The processing happens on your own computer using WebGPU and WebAssembly, two browser technologies that allow computationally intensive tasks to run locally rather than sending data to a remote service. This keeps your images private and means the tool can work offline after the initial page load. A live demo is available at inpaintweb.lxfater.com for trying it out without any installation. The frontend code is adapted from a project called cleanup.pictures, and the inpainting model is based on MI-GAN from Picsart AI Research. The project roadmap lists two features as planned but not yet complete: integration with Segment Anything for quick object selection and removal, and integration with Stable Diffusion for AI-generated replacement content. The existing history tracking and post-processing steps are listed as done. Getting the development version running requires Node.js and a terminal. You run npm install to fetch dependencies and npm run start to launch the local server. The project is open source and free to use.