Pith. sign in

REVIEW 13 cited by

Timestep Embedding Tells: It's Time to Cache for Video Diffusion Model

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2411.19108 v2 pith:CX7K2MCX submitted 2024-11-28 cs.CV

classification cs.CV
keywords modeloutputsteacachedifferencescachecachingtimesteptimesteps
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

As a fundamental backbone for video generation, diffusion models are challenged by low inference speed due to the sequential nature of denoising. Previous methods speed up the models by caching and reusing model outputs at uniformly selected timesteps. However, such a strategy neglects the fact that differences among model outputs are not uniform across timesteps, which hinders selecting the appropriate model outputs to cache, leading to a poor balance between inference efficiency and visual quality. In this study, we introduce Timestep Embedding Aware Cache (TeaCache), a training-free caching approach that estimates and leverages the fluctuating differences among model outputs across timesteps. Rather than directly using the time-consuming model outputs, TeaCache focuses on model inputs, which have a strong correlation with the modeloutputs while incurring negligible computational cost. TeaCache first modulates the noisy inputs using the timestep embeddings to ensure their differences better approximating those of model outputs. TeaCache then introduces a rescaling strategy to refine the estimated differences and utilizes them to indicate output caching. Experiments show that TeaCache achieves up to 4.41x acceleration over Open-Sora-Plan with negligible (-0.07% Vbench score) degradation of visual quality.

Discussion (0). Sign in to comment.

Forward citations

Cited by 13 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Dynamic-in-Few-Step: Unifying Dynamic Computation and Few-Step Distillation for Efficient Video Generation

    cs.CV 2026-07 conditional novelty 6.5 of 10

    Joint few-step distillation and step-specific structural pruning turns a video diffusion model into a compact Mixture-of-Models that cuts 24% extra FLOPs per step and reaches 30× speedup on Wan-14B.

  2. CachedSearch: Training-Free Cached Exploration for Test-Time Search in Video Diffusion

    cs.AI 2026-07 conditional novelty 6.0 of 10

    Using cached drafts to select the winner and regenerating only that winner captures 94.7% of best-of-8 search gain at 63% of the cost.

  3. DiTango: Cost-Effective Parallel Diffusion Generation with Selective Attention State Reuse

    cs.CV 2026-07 conditional novelty 6.0 of 10

    Selective reuse of composed attention states across denoising steps lets DiTango skip both remote KV communication and attention compute for low-contribution sequence partitions, cutting multi-GPU diffusion latency by...

  4. Kaleido: Algorithm-Hardware Co-Design for Video Diffusion Transformers by Exploiting Latent Space Correlations

    cs.AR 2026-07 conditional novelty 6.0 of 10

    A channel-wise reuse algorithm plus a reconfigurable systolic accelerator skips redundant vDiT attention and MLP computation, achieving up to 5.9x speedup and 16x energy savings.

  5. Xema: Efficient Diffusion Serving through Fine-Grained Memory Management and Auto-Configuration

    cs.DC 2026-07 conditional novelty 6.0 of 10

    Trace-guided fine-grained memory control and offline joint planning raise diffusion serving SLO attainment by up to 3.7× while cutting configuration search from hours to minutes.

  6. OmniCache: A Trajectory-Oriented Global Perspective on Training-Free Cache Reuse for Diffusion Transformer Models

    cs.CV 2025-08 unverdicted novelty 6.0 of 10

    A training-free cache-reuse scheme that spreads computation across the full diffusion trajectory and subtracts estimated noise, accelerating DiT sampling with claimed competitive quality.

  7. SADA: Stability-guided Adaptive Diffusion Acceleration

    cs.LG 2025-07 conditional novelty 6.0 of 10

    SADA accelerates ODE-based generative model sampling by adaptively combining step skipping and token pruning through a stability criterion, giving about 1.8 times speedup with minor fidelity loss.

  8. TaoMate: Anchor-Guided Memory Bridging Evolving and Reference States for Real-Time Audio-Video Digital Human Generation

    cs.CV 2026-07 conditional novelty 5.0 of 10

    Anchor-guided fixed-capacity multimodal memory plus stage-parallel few-step denoising stabilizes long-form causal audio-video digital humans at real-time speed.

  9. OmniCache: Multidimensional Hierarchical Feature Caching For Diffusion Models

    cs.CV 2026-07 conditional novelty 5.0 of 10

    Training-free hierarchical feature caching across token, frame, block, and layer axes cuts diffusion inference latency up to 35% while preserving quality better than averaging-based token merging.

  10. FSampler: Training Free Acceleration of Diffusion Sampling via Epsilon Extrapolation

    cs.LG 2025-11 conditional novelty 4.0 of 10

    FSampler accelerates diffusion sampling by substituting finite-difference extrapolations of epsilon for model calls on selected steps, reducing NFE by 15-25% at SSIM 0.95-0.99.

  11. DriveGen3D: Boosting Feed-Forward Driving Scene Generation with Efficient Video Diffusion

    cs.CV 2025-10 conditional novelty 4.0 of 10

    DriveGen3D makes long driving-video synthesis and 3D scene reconstruction practical by caching only the conditional diffusion branch, quantizing cross-view attention, and fusing temporal context into a feed-forward Ga...

  12. Predictive Feature Caching for Training-free Acceleration of Molecular Geometry Generation

    cs.LG 2025-10 conditional novelty 4.0 of 10

    Predictive feature caching, borrowed from image diffusion, speeds up molecular flow-matching generation by 2-3x at near-matched quality by forecasting hidden features instead of recomputing them.

  13. Token-Operations-Oriented Inference Optimization Techniques for Large Models

    cs.SE 2026-06 unverdicted novelty 3.0 of 10

    The paper introduces a four-layer technical architecture for token-operations-oriented inference optimization in large models and reviews key technologies and industry status at each layer.

Pith tools