LSGQuant: Layer-Sensitivity Guided Quantization for One-Step Diffusion Real-World Video Super-Resolution

One-step diffusion models have demonstrated a promising capability and fast inference in video super-resolution (VSR) for real-world applications. Nevertheless, the substantial model size and high computational cost of Diffusion Transformers (DiTs) limit downstream applications. While low-bit quantization is a common approach for model compression, the effectiveness of quantized models is challenged by the high dynamic range of input latent and diverse layer behaviors. To address these challenges, we introduce LSGQuant, a layer-sensitivity-guided quantization approach for one-step diffusion-based real-world VSR. Our method incorporates a Dynamic Range Adaptive Quantizer (DRAQ) to fit video token activations. Furthermore, we estimate layer sensitivity and implement a Variance-Oriented Layer Training Strategy (VOLTS) by analyzing layer-wise statistics in calibration. We also introduce Quantization-Aware Optimization (QAO) to jointly refine the quantized branch and a retained high-precision branch. Extensive experiments demonstrate that our method has nearly the performance of the original model with full-precision and significantly exceeds existing quantization techniques. All models and code are available at https://github.com/zhengchen1999/LSGQuant.

Making low-resolution videos look sharp and clear on your phone often requires powerful AI models. These models are too large and slow to run efficiently on everyday devices, limiting their use for real-time video super-resolution. We proposed a new method called LSGQuant that shrinks the AI model’s size using a technique called "quantization" by using simpler numbers for calculations. LSGQuant intelligently identifies which parts of the model are most sensitive to this compression and protects them. It also adapts to the wide range of visual information in videos, reducing errors that typically occur during calculation. Our approach compresses the video super-resolution model to use 70% less memory and computation while maintaining almost the same quality as the original and much larger model. This makes it practical to run high-quality video super-resolution directly on smartphones or other edge devices, enabling clearer video calls, better streaming, and more detailed playback.