ACID: Adaptive Caching for vIDeo generation

2026-07-14

AI Summary

This work addresses the inherent speed–quality trade-off in video diffusion models caused by fixed-threshold caching strategies, which fail to adapt to the dynamic nature of denoising steps. The authors propose ACID, a lightweight, training-free adaptive caching framework that introduces a dynamic threshold mechanism for the first time. By monitoring the rate of change in drift signals, ACID identifies critical denoising steps and applies low thresholds only there, while aggressively caching in non-critical steps. This approach breaks the conventional trade-off boundary without requiring model modifications or retraining, and is compatible with existing methods such as TeaCache, EasyCache, and DiCache. Evaluated on three open-source models including HunyuanVideo, ACID achieves up to 38% additional speedup over conservative fixed-threshold baselines (2.16× total acceleration) with negligible quality degradation (PSNR < 0.3 dB, SSIM/LPIPS < 0.01).

Abstract

Video diffusion models produce high-quality generations but remain slow at inference due to their sequential denoising procedure. Caching-based acceleration methods address this by reusing intermediate model outputs: leading dynamic approaches such as TeaCache, EasyCache, and DiCache accumulate a drift signal and skip expensive model evaluations when accumulated drift stays below a fixed threshold τ. This threshold controls an apparent tradeoff - raising it yields faster generation at the cost of visual quality, while lowering it preserves quality but sacrifices speed. We show this tradeoff is not fundamental; it is an artifact of holding τ constant throughout denoising. We identify the existence of critical steps - timesteps where the drift signal changes rapidly - and show that applying a low threshold selectively at these steps while caching aggressively elsewhere recovers most of the quality of conservative caching at substantially higher inference speeds. Building on this insight, we propose ACID, a lightweight, training-free wrapper that monitors the rate of change of each method's existing drift signal to dynamically switch between a low and a high threshold. ACID is signal-agnostic and modular: it requires no retraining and plugs directly into existing dynamic caching methods without modifying their core mechanisms. Evaluated across three caching methods (TeaCache, EasyCache, DiCache) and three open-source video diffusion models (HunyuanVideo, Wan 2.1, CogVideoX), ACID consistently expands the Pareto frontier of visual quality versus inference speed beyond what any fixed threshold achieves. In particular, on TeaCache and HunyuanVideo, ACID achieves up to 2.16x speedup over the no-caching baseline, and up to 38% additional speedup over the conservative fixed-threshold baseline with negligible (<0.3 dB PSNR, <0.01 SSIM, <0.01 LPIPS) quality degradation.

Problem

• video diffusion models
• inference acceleration
• caching
• quality-speed tradeoff
• denoising

Innovation

• adaptive caching
• video diffusion models
• dynamic thresholding
• inference acceleration
• drift signal