Inpaint-web is a browser-based image editing tool that performs inpainting and image upscaling using WebGPU and WebAssembly.
The tool solves the problem of removing unwanted objects from images and enhancing image resolution entirely within the browser, without requiring server-side processing. It uses WebGPU for GPU acceleration and WebAssembly to run machine learning models directly in the client, keeping all computation local to the user's machine. Users can paint over areas they want to remove or modify, and the tool reconstructs those regions. The project also supports super-resolution upscaling to enhance image clarity.
Developers should choose this tool if they need client-side image manipulation without external dependencies or API calls. It suits projects where privacy is important, since processing happens entirely in the browser. The tool is free and open-source, making it accessible for both personal and commercial use. The README acknowledges that frontend code was modified from cleanup.pictures and credits the MI-GAN model for the underlying inpainting capability.
The project maintains an active roadmap with completed features including image modification history, model optimization, and integrated post-processing. Planned additions include integration with Segment Anything for automated selection and removal, and Stable Diffusion for image replacement capabilities. Development appears focused on expanding functionality while the maintainer notes time constraints due to other business activities.
