MagCache: Fast Video Generation with Magnitude-Aware Cache
This work addresses the need for faster video generation without extensive calibration, offering a robust solution for applications requiring efficient high-quality video synthesis.
The paper tackles the problem of accelerating video diffusion models by introducing MagCache, a magnitude-aware cache that adaptively skips timesteps based on a unified magnitude law, achieving 2.10x-2.68x speedups on models like Open-Sora while preserving visual fidelity with superior LPIPS, SSIM, and PSNR scores.
Existing acceleration techniques for video diffusion models often rely on uniform heuristics or time-embedding variants to skip timesteps and reuse cached features. These approaches typically require extensive calibration with curated prompts and risk inconsistent outputs due to prompt-specific overfitting. In this paper, we introduce a novel and robust discovery: a unified magnitude law observed across different models and prompts. Specifically, the magnitude ratio of successive residual outputs decreases monotonically, steadily in most timesteps while rapidly in the last several steps. Leveraging this insight, we introduce a Magnitude-aware Cache (MagCache) that adaptively skips unimportant timesteps using an error modeling mechanism and adaptive caching strategy. Unlike existing methods requiring dozens of curated samples for calibration, MagCache only requires a single sample for calibration. Experimental results show that MagCache achieves 2.10x-2.68x speedups on Open-Sora, CogVideoX, Wan 2.1, and HunyuanVideo, while preserving superior visual fidelity. It significantly outperforms existing methods in LPIPS, SSIM, and PSNR, under similar computational budgets.