RVRT is a transformer-based architecture published at NeurIPS 2022 that addresses low-level video processing tasks including super-resolution, deblurring, and denoising. The model employs guided deformable attention mechanisms and recurrent structures to process video frames efficiently while maintaining temporal consistency. It provides pretrained models and achieves state-of-the-art results on standard benchmarks including REDS, Vimeo90K, GoPro, and DAVIS datasets.