PromptTea: Let Prompts Tell TeaCache the Optimal Threshold

• URL: http://arxiv.org/abs/2507.06739v1
• Date: Wed, 09 Jul 2025 10:53:05 GMT
• Title: PromptTea: Let Prompts Tell TeaCache the Optimal Threshold
• Authors: Zishen Huang, Chunyu Yang, Mengyuan Ren,
• Abstract summary: A common acceleration strategy involves reusing model outputs via caching mechanisms at fixed intervals.<n>We propose Prompt-Complexity-Aware (PCA) caching, a method that automatically adjusts reuse thresholds based on scene complexity estimated directly from the input prompt.
• Score: 1.0665410339553834
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Despite recent progress in video generation, inference speed remains a major bottleneck. A common acceleration strategy involves reusing model outputs via caching mechanisms at fixed intervals. However, we find that such fixed-frequency reuse significantly degrades quality in complex scenes, while manually tuning reuse thresholds is inefficient and lacks robustness. To address this, we propose Prompt-Complexity-Aware (PCA) caching, a method that automatically adjusts reuse thresholds based on scene complexity estimated directly from the input prompt. By incorporating prompt-derived semantic cues, PCA enables more adaptive and informed reuse decisions than conventional caching methods. We also revisit the assumptions behind TeaCache and identify a key limitation: it suffers from poor input-output relationship modeling due to an oversimplified prior. To overcome this, we decouple the noisy input, enhance the contribution of meaningful textual information, and improve the model's predictive accuracy through multivariate polynomial feature expansion. To further reduce computational cost, we replace the static CFGCache with DynCFGCache, a dynamic mechanism that selectively reuses classifier-free guidance (CFG) outputs based on estimated output variations. This allows for more flexible reuse without compromising output quality. Extensive experiments demonstrate that our approach achieves significant acceleration-for example, 2.79x speedup on the Wan2.1 model-while maintaining high visual fidelity across a range of scenes.

Related papers

• Less is Enough: Training-Free Video Diffusion Acceleration via Runtime-Adaptive Caching [57.7533917467934]
  EasyCache is a training-free acceleration framework for video diffusion models.<n>We conduct comprehensive studies on various large-scale video generation models, including OpenSora, Wan2.1, and HunyuanVideo.<n>Our method achieves leading acceleration performance, reducing inference time by up to 2.1-3.3$times$ compared to the original baselines.
  arXiv  Detail & Related papers  (2025-07-03T17:59:54Z)
• MagCache: Fast Video Generation with Magnitude-Aware Cache [91.51242917160373]
  We introduce a novel and robust discovery: a unified magnitude law observed across different models and prompts.<n>We introduce a Magnitude-aware Cache (MagCache) that adaptively skips unimportant timesteps using an error modeling mechanism and adaptive caching strategy.<n> Experimental results show that MagCache achieves 2.1x and 2.68x speedups on Open-Sora and Wan 2.1, respectively.
  arXiv  Detail & Related papers  (2025-06-10T17:59:02Z)
• Model Reveals What to Cache: Profiling-Based Feature Reuse for Video Diffusion Models [41.11005178050448]
  ProfilingDiT is a novel adaptive caching strategy that explicitly disentangles foreground and background-focused blocks.<n>Our framework achieves significant acceleration while maintaining visual fidelity across comprehensive quality metrics.
  arXiv  Detail & Related papers  (2025-04-04T03:30:15Z)
• Exposure Bias Reduction for Enhancing Diffusion Transformer Feature Caching [7.393824353099595]
  Diffusion Transformer (DiT) has exhibited impressive generation capabilities but faces great challenges due to its high computational complexity.<n>We analyze the impact of caching on the SNR of the diffusion process.<n>We introduce EB-Cache, a joint cache strategy that aligns the Non-exposure bias.
  arXiv  Detail & Related papers  (2025-03-10T09:49:18Z)
• QuantCache: Adaptive Importance-Guided Quantization with Hierarchical Latent and Layer Caching for Video Generation [84.91431271257437]
  Diffusion Transformers (DiTs) have emerged as a dominant architecture in video generation.<n>DiTs come with significant drawbacks, including increased computational and memory costs.<n>We propose QuantCache, a novel training-free inference acceleration framework.
  arXiv  Detail & Related papers  (2025-03-09T10:31:51Z)
• Timestep Embedding Tells: It's Time to Cache for Video Diffusion Model [55.64316746098431]
  Timestep Embedding Aware Cache (TeaCache) is a training-free caching approach that estimates and leverages the fluctuating differences among model outputs across timesteps.<n>TeaCache achieves up to 4.41x acceleration over Open-Sora-Plan with negligible degradation of visual quality.
  arXiv  Detail & Related papers  (2024-11-28T12:50:05Z)
• FasterCache: Training-Free Video Diffusion Model Acceleration with High Quality [58.80996741843102]
  FasterCache is a training-free strategy designed to accelerate the inference of video diffusion models with high-quality generation.<n>We show that FasterCache can significantly accelerate video generation while keeping video quality comparable to the baseline.
  arXiv  Detail & Related papers  (2024-10-25T07:24:38Z)
• Accelerating Deep Learning Classification with Error-controlled Approximate-key Caching [72.50506500576746]
  We propose a novel caching paradigm, that we named approximate-key caching. While approximate cache hits alleviate DL inference workload and increase the system throughput, they however introduce an approximation error. We analytically model our caching system performance for classic LRU and ideal caches, we perform a trace-driven evaluation of the expected performance, and we compare the benefits of our proposed approach with the state-of-the-art similarity caching.
  arXiv  Detail & Related papers  (2021-12-13T13:49:11Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.