README.md

---
title: SwiftVR
emoji: 🎬
colorFrom: indigo
colorTo: yellow
sdk: gradio
sdk_version: "6.19.0"
app_file: app.py
short_description: Real-time one-step generative video restoration
python_version: "3.10"
startup_duration_timeout: 1h
---

# SwiftVR: Real-Time One-Step Generative Video Restoration

This Space demonstrates [SwiftVR](https://github.com/H-oliday/SwiftVR), the first generative video restoration model to reach real-time 1080p streaming on a consumer-grade GPU.

## How it works

Upload a low-quality video and SwiftVR will restore and upscale it in a single forward pass — no iterative diffusion sampling needed. The model uses a one-step DiT (Diffusion Transformer) with a Restoration-aware Autoencoder (ReAE) and causal chunk-wise streaming.

## Links

- [Paper (arXiv 2606.09516)](https://arxiv.org/abs/2606.09516)
- [GitHub](https://github.com/H-oliday/SwiftVR)
- [Model on Hugging Face](https://huggingface.co/H-oliday/SwiftVR)