RIFE implements a deep learning model for video frame interpolation by estimating intermediate optical flow between consecutive frames. The model uses a lightweight architecture to achieve 30+ FPS for 2X 720p interpolation on a 2080Ti GPU, supporting arbitrary-timestep interpolation. It builds on a published ECCV 2022 paper and includes optimizations for anime scenes and post-processing of diffusion-generated videos.