Parallel sampling speeds up diffusion image generation with minimal quality loss
ParaLin++: Accelerating Parallel Diffusion Integrator via Intrinsic Partially Linear Structure
Computer Vision and Pattern Recognition
Summary
Generating images with diffusion models usually takes many steps and can be slow. People try to speed this up by using fewer steps, but this often reduces image quality when tested on harder benchmarks. The authors found that parallel sampling, which changes how steps are done without skipping, keeps image quality high while running much faster. They made a new method called ParaLin++ that uses this idea and showed it works well on several image datasets and models.
What this means in practice
- •For graphics software developers: Speed up image generation in creative tools by integrating ParaLin++ to keep high visual quality with fewer iterations.
- •For cloud ai service providers: Reduce GPU time costs by deploying ParaLin++ for faster diffusion model inference without noticeable drop in output quality.$Commercial implications: Enables more efficient paid image generation services by cutting compute time while maintaining client satisfaction.
Authors
Jianrong Lu, Zhiyu Zhu, Hui Liu, Junhui Hou
Abstract
Training-free few-step samplers are often described as nearly quality-lossless under conventional FID/CLIP-style evaluation. It remains unclear whether that conclusion transfers to more realistic and complex generation-quality benchmarks. We first systematize the quality metrics used to evaluate training-free sampling acceleration for diffusion and flow-matching models. We find that representative training-free step-shortening methods often remain nearly lossless under FID/CLIP-class metrics, yet can suffer substantial quality degradation when evaluated with semantic-composition, dense-instruction, and long-text-rendering benchmarks such as GenEval, DPG-Bench, and LongTextBench. Parallel sampling is a notable exception: by retaining the full-step discretization and removing only the solver's temporal dependency, it can preserve near-lossless quality while accelerating sampling. Building on this observation, we propose \textit{ParaLin++}, a step-parity parallel sampler that preserves the full-step discretization while accelerating the solver iterations. Experiments across Stable Diffusion-v2, LSUN Church, FLUX.1-dev, and Ming-Flash-Omni 2.0 show that ParaLin++ preserves the full-step Euler quality profile at $2.6\times$--$4.8\times$, $1.68\times$ on FLUX, and $1.21\times$ on Ming. Project page: https://paralinplus.github.io/.