Video2X is a C++ rewrite of a video enhancement framework that applies machine learning models to upscale videos and interpolate frames. It supports multiple neural network architectures including Real-ESRGAN, Real-CUGAN, RIFE, and Anime4K, running them via ncnn and Vulkan for GPU acceleration. The tool operates in two modes: filtering (upscaling) and frame interpolation, processing videos without requiring additional disk space during processing.