SwiftVR: Real-Time One-Step Generative Video Restoration

2026-06-08

AI Summary

This work addresses the challenges of single-frame latency and GPU memory bottlenecks in real-time, high-resolution video inpainting for live streaming scenarios by proposing SwiftVR, a one-step generative video inpainting framework grounded in a causal chunking protocol. SwiftVR introduces a mask-free shifted-window self-attention mechanism coupled with an inpainting-aware lightweight autoencoder, enabling efficient inference under standard dense attention—without requiring custom sparse kernels or cyclic shifts. The method achieves real-time 1080p inpainting at 26 FPS on consumer-grade GPUs (e.g., RTX 5090) and scales to 2560×1440@31 FPS and 4K@14 FPS on H100 accelerators, significantly outperforming existing diffusion-based approaches while maintaining low latency, minimal memory footprint, and high reconstruction quality.

Abstract

Real-time video restoration (VR) for live streams requires high-resolution outputs under strict per-frame latency constraints. Existing one-step diffusion-based VR models remain difficult to deploy on consumer-grade GPUs due to two main bottlenecks: quadratic spatial attention at high resolutions and the latency-memory overhead of large video autoencoders. We present SwiftVR, a streaming one-step generative VR framework that reduces both bottlenecks under a causal chunk-wise protocol. For attention, mask-free shifted-window self-attention gathers each spatial window into a dense tensor via deterministic indexing, keeping all attention calls on the dense scaled dot-product attention path without masks, cyclic shifts, padding, or hardware-specific sparse kernels. Because SwiftVR uses only standard dense SDPA calls, the trained model transfers to consumer GPUs without retraining or custom kernels. For autoencoding, a lightweight Restoration-aware Autoencoder enables fast chunk-wise decoding while preserving reconstruction quality. On a single H100, SwiftVR sustains 31~FPS at 2560x1440 and 14~FPS at 3840x2160, whereas all compared diffusion-based VR baselines exceed the memory limit at 4K. On a consumer RTX~5090, SwiftVR reaches 26~FPS at 1920x1080. To our knowledge, SwiftVR is the first generative VR model to achieve real-time 1080p streaming on a consumer-grade GPU, while attaining strong no-reference perceptual quality with lower inference cost. Project is available at https://h-oliday.github.io/SwiftVR.

Problem

• video restoration
• real-time
• diffusion models
• consumer GPU
• high-resolution

Innovation

• one-step diffusion
• real-time video restoration
• mask-free shifted-window attention
• restoration-aware autoencoder
• consumer GPU deployment