eutropicai/final2x

Final2x is a desktop application that takes a lower-resolution image and produces a higher-resolution version using AI-based super-resolution models. It runs on Mac, Windows, and Linux, packaged as a native desktop app via the Electron framework, so users get a graphical window rather than a command line. The processing backend is a separate component called Final2x-core, which handles the actual computation. Users load an image through the interface and choose a scale factor, the backend then runs the upscaling using pre-trained AI models. Version 4.0 introduced a new backend called cccv that adds support for custom models, meaning users can supply their own trained networks beyond the built-in options. A demo repository shows how remote custom models can be configured. Version 3.0 expanded hardware support to include Nvidia's 50-series graphics cards. Installation differs by platform. On Windows, package managers like winget and scoop provide a convenient install path, though those channels may lag behind the latest official release. On Mac, first-time setup requires running two terminal commands to bypass Gatekeeper, the operating system's security check that blocks applications not distributed through Apple's official channels. On Linux, users need to install Python 3.9 or newer and PyTorch 2.0 or newer before the app will run, plus two additional system libraries. The desktop shell and interface are written in TypeScript. The UI layer uses a component library called naive-ui, and the build tooling is electron-vite, which packages the web-based interface into a native app bundle. The Final2x-core backend, also available as a standalone command-line tool installable via pip, is what bridges the desktop interface to the underlying PyTorch model inference. Developers working on the project can run the core separately without the Electron wrapper. The project is released under the BSD 3-Clause license.