STAR: Spatial-Temporal Augmentation with Text-to-Video Models for Real-World Video Super-Resolution

Image diffusion models have been adapted for real-world video super-resolution to tackle over-smoothing issues in GAN-based methods. However, as they are trained on static images, they struggle to maintain temporal consistency, limiting their ability to capture temporal dynamics effectively. Introducing text-to-video (T2V) models into video super-resolution for improved temporal modeling is straightforward. However, two key challenges remain: artifacts caused by complex degradations in real-world scenarios and compromised fidelity due to the strong generative capacity of powerful T2V models (e.g., CogVideoX-5B). To enhance the spatio-temporal quality of restored videos, we introduce STAR (Spatial-Temporal Augmentation with T2V models for Real-world video super-resolution), a novel approach that leverages T2V models for real-world video super-resolution, achieving realistic spatial details and robust temporal consistency. Specifically, we enhance vanilla T2V models by incorporating Local Information Enhancement to enrich local details and mitigate degradation artifacts. Moreover, we propose a Dynamic Frequency (DF) Loss to reinforce fidelity, guiding the model to focus on different frequency components across diffusion steps. Extensive experiments demonstrate STAR outperforms state-of-the-art methods on both synthetic and real-world datasets.