InfVSR: Breaking Length Limits of Generic Video Super-Resolution
Researchers have introduced InfVSR, a novel approach to video super-resolution (VSR) designed to overcome the limitations of processing long video sequences. By reformulating VSR as an autoregressive one-step diffusion process, InfVSR enables efficient streaming inference and maintains temporal consistency. The method adapts a pretrained diffusion model with a causal structure and distills the diffusion process into a single step, achieving significant speed-ups and improved quality on a new benchmark for extended videos. AI
IMPACT InfVSR offers a significant speed-up for video super-resolution, potentially enabling new applications for processing long-form video content.