FastVR is a video restoration framework that processes 1080p video at 11 FPS on a single H20 GPU, making it suitable for real-time applications. It uses a one-step to avoid multi-step sampling costs, a lightweight VAE to reduce encoding/decoding overhead, and chunk-wise causal attention to limit attention complexity. The model achieves strong restoration quality and temporal consistency, outperforming prior diffusion-based methods in efficiency while maintaining competitive visual fidelity.
One-step diffusion framework restores 1080p video in real time with efficient streaming attention
FastVR combines a lightweight VAE with chunk-wise causal attention to cut the two dominant costs of diffusion video restoration while maintaining state-of-the-art quality.
Chinese Tech
Xiaoxu Chen · Qin Yang · Haoran Bai · Sibin Deng · Ying Chen
Alibaba Group · Xidian University
Research Digest··3 min read
In one line
FastVR restores high-resolution video at 11 FPS using a one-step diffusion model with a lightweight VAE and chunk-wise causal attention.
Before you read this
Three ideas this paper builds ondiffusion modelA type of generative model that learns to reverse a gradual noising process to create data.199 papers →diffusion transformerA neural network architecture combining transformer layers with diffusion processes for high-quality image and video generation.76 papers →sparse attentionSelectively attending to only the most relevant positions or tokens to reduce computational overhead.51 papers →
Read it as
Chen et al. present FastVR, a one-step diffusion framework for streaming video restoration that processes 1080p video at 11 FPS on a single H20 GPU.
FastVR builds on a one-step , eliminating the iterative denoising loop of multi-step baselines such as Vivid-VR.
How it works
stage by stageThe Daily Edition
The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.