WebSR: Real-time AI Upscaling of videos and images on the web (github.com)

WebSR has launched a new, free, and open-source JavaScript library that enables real-time AI upscaling of videos and images directly within web browsers. Built with WebGPU compute shaders, the library supports multiple network sizes—Small, Medium, and Large—allowing developers to balance performance and quality. Notably, it comes with pre-trained models optimized for different content types, including animations, real-life visuals, and 3D graphics. While still in development, WebSR aims to set a new standard for super-resolution techniques on the web, addressing the high computational demands that have historically hindered the adoption of this technology for video streaming. This initiative is significant for the AI/ML community as it opens up practical applications for super-resolution in real-time video scenarios, such as WebRTC and low-bandwidth situations. Developers can customize their models for specific content types through an advanced training pipeline included in the project. With features like offscreen processing to keep the main thread responsive and the ability to dynamically switch networks, WebSR is positioned to enhance user experiences in web applications that require video quality improvements, paving the way for future innovations in browser-based AI solutions.