video2x

Video2X is a machine learning-based video super resolution and frame interpolation framework. Version 6.0.0 represents a complete rewrite of the project in C/C++, designed for a faster and more efficient architecture with cross-platform support for Windows and Linux. Key capabilities include: - Two processing modes: filtering (upscaling) and frame interpolation. - Support for Anime4K v4 and custom MPV-compatible GLSL shaders. - Support for Real-ESRGAN, Real-CUGAN, and RIFE (all models) via ncnn and Vulkan. - Zero additional disk space required during processing, only space for the final output. - A graphical user interface (GUI) and installer for Windows, supporting multiple languages including English, Simplified Chinese, Japanese, Portuguese, French, and German. Hardware requirements include a CPU with AVX2 support and a GPU that supports Vulkan. The project is available as precompiled binaries for Windows and Linux, an AppImage for other Linux distributions, a container image on GitHub Container Registry for Linux and macOS deployment, and a free Google Colab notebook for users without a powerful GPU. The project is licensed under GNU AGPL version 3 and depends on several open-source projects including FFmpeg, Tencent's ncnn, Anime4K, realcugan-ncnn-vulkan, rife-ncnn-vulkan, and Real-ESRGAN-ncnn-vulkan.