Analyzing and Improving Fast Sampling of Text-to-Image Diffusion Models

Zhenyu Zhou; Defang Chen; Siwei Lyu; Chun Chen; Can Wang

cs.CVFeb 28, 2026

Analyzing and Improving Fast Sampling of Text-to-Image Diffusion Models

Authors: Zhenyu Zhou, Defang Chen, Siwei Lyu, Chun Chen, Can Wang

Organizations: State Key Laboratory of Blockchain and Data Security, Zhejiang University · Hangzhou High-Tech Zone (Binjiang) Institute of Blockchain and Data Security · University of California, Berkeley · State University of New York at Buffalo

Abstract

Text-to-image diffusion models have achieved unprecedented success but still struggle to produce high-quality results under limited sampling budgets. Existing training-free sampling acceleration methods are typically developed independently, leaving the overall performance and compatibility among these methods unexplored. In this paper, we bridge this gap by systematically elucidating the design space, and our comprehensive experiments identify the sampling time schedule as the most pivotal factor. Inspired by the geometric properties of diffusion models revealed through the Frenet-Serret formulas, we propose constant total rotation schedule (TORS), a scheduling strategy that ensures uniform geometric variation along the sampling trajectory. TORS outperforms previous training-free acceleration methods and produces high-quality images with 10 sampling steps on Flux.1-Dev and Stable Diffusion 3.5. Extensive experiments underscore the adaptability of our method to unseen models, hyperparameters, and downstream applications. Code is available at https://github.com/zju-pi/TORS.

Figures & tables

Figure 1: Text-to-image generation using Flux.1-Dev [ 1 ] . Our proposed TORS attains image quality comparable to the 50-step baseline using only 10 steps.

Panel of Figure 2 — Figure 2: (a) Diffusion sampling follows a predefined outer schedule $\{t_{n}\}_{n=0}^{N}$ , while an inner schedule determines whether each step performs computation or feature reuse. A sample $\mathbf{x}_{t_{n+1}}$ transitions from $t_{n+1}$ to $t_{n}$ with the solver’s update rule, utilizing stored historical velocities. Various types of cache objects can be selected, and a feature predictor is employed to estimate the current feature $\Delta_{t_{n}}$ . This perspective also applies to U-Net architectures. (b) Four types of cache objects: velocity cache (V. cache), transformer cache (T. cache), block cache (B. cache), and operation cache (O. cache).

Panel of Figure 3 — Figure 3: Comprehensive experiments examining the impact of each acceleration method on sampling acceleration. The experiments are conducted on Flux.1-Dev [ 1 ] . The outer schedule is identified as the most influential factor (see the vertical scale range).

Figure 4: Text-to-image generation with Flux.1-Dev [ 1 ] using the default uniform outer schedule. As the number of sampling steps increases, the image structure continues to change and only stabilizes after around 30 steps.

Panel of Figure 5 — Figure 5: We extend the insights of sampling regularity from diffusion models to flow-based models and propose TORS by utilizing the geometric properties of sampling trajectories. (a) Visualization of 100 sampling trajectories generated by Flux.1-Dev reveals a strong trajectory regularity. All sampling trajectories starting from Gaussian noise are uniformly shifted to the origin. (b) A 10-step example of our proposed TORS, which ensures a constant total rotation change $\int_{0}^{S}|\omega(s)|\mathrm{d}s$ along the sampling trajectory. (c) TORS yields considerably faster structural convergence compared with the uniform outer schedule ( Figure 4 ).

Table 1: Results on text-to-image generation. Our proposed TORS achieves state-of-the-art results among sampling acceleration methods.

Model	Steps	Single Object	Two Object	Counting	Colors	Attribute Binding	Position	Overall $\uparrow$
Baseline	50	0.99	0.81	0.69	0.81	0.41	0.20	0.65
Baseline	10	0.96	0.71	0.59	0.69	0.28	0.20	0.57
TORS (ours)	10	1.00	0.82	0.67	0.80	0.49	0.22	0.67

Table 2: GenEval evaluation on Flux.1-dev.

Table 9

Table 5: Ablation studies evaluating the robustness of TORS schedules on prompt distribution, sample sizes and step counts.

Method	Structure	Background Preservation				CLIP Similarity
Method	Dist. ${}_{\times 10^{3}}$ $\downarrow$	PSNR $\uparrow$	LPIPS ${}_{\times 10^{3}}$ $\downarrow$	MES ${}_{\times 10^{4}}$ $\downarrow$	SSIM ${}_{\times 10^{2}}$ $\uparrow$	Whole $\uparrow$	Edit $\uparrow$
Baseline $N=50$	47.52	28.33	59.32	90.17	91.67	25.71	22.65
Baseline $N=10$	65.29	25.05	75.99	112.10	87.69	25.64	22.59
TORS $N=10$	52.46	27.59	64.87	90.90	90.37	25.76	22.70

Table 6: Performance of image editing evaluated on PIE-Bench using Flux.1-Kontext.

Figure 9: Qualitative results of image editing using Flux.1-Kontext.

Panel of Figure 10 — Figure 10: Compatibility evaluation on acceleration methods using Image Reward. The performance can be enhanced by integrating appropriate acceleration strategies.

Appendix figures & tables9 assets

Supplementary material from the paper’s appendix.

Appendix

\int_{0}^{s_{i}}|\boldsymbol{\omega}(s)|ds=\frac{i}{N}\Theta_{[0,S]},\quad\text{for }i=0,\dots,N.

Appendix

Algorithm 1 Generate TORS Schedule

Cache Object	Latency (s) $\downarrow$			VRAM (GB) $\downarrow$
Cache Object	$O_{F}=1/2/3$			$O_{F}=1/2/3$
V. cache	9.85 /	– /	–	36.32 /	– /	–
T. cache	9.90 /	9.92 /	9.94	36.38 /	36.44 /	38.61
B. cache	10.03 /	10.41 /	10.64	37.94 /	39.56 /	41.17
O. cache	10.47 /	10.87 /	11.19	38.48 /	40.63 /	42.78

Appendix

Table 7: Latency and VRAM requirements on Flux.1-Dev with a batch size of 1. All statistics are measured on a single A6000 GPU. Velocity cache requires 38.88 GB, 41.44 GB and 44.00 GB of VRAM for batch sizes of 2, 3 and 4, respectively.

Figure 16: Text-to-image generation with Stable Diffusion 3.5 medium [ 6 ] using the default uniform outer schedule.

Figure 18: Additional qualitative results using Flux.1-Dev [ 1 ] .

Figure 19: Additional qualitative results using Stable Diffusion 3.5 medium [ 6 ] .

Explore similar work

Optimize Your Sampling: Tuned Diffusion Sampling with Bayesian Optimization
Aug 18, 2026Travis Zhang, Christian Belardi, Justin Lovelace +4Diffusion Timestep Scheduling Few-Step Diffusion Sampling
Cornell University
Multi-Resolution Flow Matching: Training-Free Diffusion Acceleration via Staged Sampling
Jul 2, 2026Xingyu Zheng, Xianglong Liu, Yifu Ding +4Diffusion Model Acceleration Diffusion Model Inference Acceleration
State Key Laboratory of Complex & Critical Software Environment, Beihang University · School of Computer Science and Engineering, Beihang University · Nanyang Technological University +5
Learning Sampling Parameters for Diffusion Models
Jul 26, 2026Arisrei Lim, Yossi GandelsmanDiffusion Model Guidance Diffusion Model Sampling
Reve

Aug 18, 2026J/K move · Enter open · S save

Optimize Your Sampling: Tuned Diffusion Sampling with Bayesian Optimization

Travis Zhang, Christian Belardi, Justin Lovelace, Jin Peng Zhou+3

Cornell University

Diffusion Timestep Scheduling Few-Step Diffusion Sampling Diffusion Model Inference Acceleration T2I Diffusion Models

Open paper AskPDF arXiv

Sampling from a diffusion model typically requires many forward passes through a large neural network, making generation computationally expensive. While much work has focused on efficient solvers and samplers, comparatively little attention has been paid to selecting the sampling timesteps themselves. A recent line of work optimizes theoretically derived surrogates for sample quality rather than the quality metric itself. We propose Optimizing Your Sampling (OYS), which instead treats timestep selection as a black-box optimization problem, optimizing the target metric directly with Bayesian optimization. OYS outperforms both the default schedules and those of Align Your Steps on text-to-image generation, and improves over the default schedules on inpainting and other image tasks, in both quantitative and human evaluations. OYS requires no additional training, is applicable even to distilled models, and improves both simple and sophisticated samplers such as Euler and DPM-Solver++. A 5-step OYS schedule retains 89%-94% of the quality of a 50-step schedule while reducing inference cost by 10x.

Abstract

Explore similar work

Optimize Your Sampling: Tuned Diffusion Sampling with Bayesian Optimization

Multi-Resolution Flow Matching: Training-Free Diffusion Acceleration via Staged Sampling

Learning Sampling Parameters for Diffusion Models