TaoCache通过固定点预测加速视频生成，保持高分辨率结构一致性。
TaoCache: Structure-Maintained Video Generation Acceleration
- 以噪声输出的固定点视角替代残差缓存，专注后期去噪阶段
- 在相同加速比下，LPIPS、SSIM、PSNR均优于现有缓存方法
- 无需训练，可与PAB、TeaCache等技术无缝集成，适合高效视频生成场景
现有的基于缓存的视频扩散模型加速方法主要跳过早期或中期去噪步骤，常导致与全步数生成相比出现结构偏差，影响指令遵循和角色一致性。我们提出TaoCache，一种无需训练、即插即用的缓存策略，不采用残差式缓存，而是从固定点视角预测模型的噪声输出，特别适用于后期去噪阶段。通过校准连续噪声增量之间的余弦相似度与范数比，TaoCache在实现激进跳步的同时保留了高分辨率结构。该方法与金字塔注意力广播（PAB）和TeaCache等互补加速技术正交，可无缝集成至DiT框架。在Latte-1、OpenSora-Plan v110和Wan2.1上，TaoCache在相同加速比下显著提升了视觉质量（LPIPS、SSIM、PSNR），优于以往缓存方法。
原文摘要 · Abstract (English)
Existing cache-based acceleration methods for video diffusion models primarily skip early or mid denoising steps, which often leads to structural discrepancies relative to full-timestep generation and can hinder instruction following and character consistency. We present TaoCache, a training-free, plug-and-play caching strategy that, instead of residual-based caching, adopts a fixed-point perspective to predict the model's noise output and is specifically effective in late denoising stages. By calibrating cosine similarities and norm ratios of consecutive noise deltas, TaoCache preserves high-resolution structure while enabling aggressive skipping. The approach is orthogonal to complementary accelerations such as Pyramid Attention Broadcast (PAB) and TeaCache, and it integrates seamlessly into DiT-based frameworks. Across Latte-1, OpenSora-Plan v110, and Wan2.1, TaoCache attains substantially higher visual quality (LPIPS, SSIM, PSNR) than prior caching methods under the same speedups.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。