Low-motion scenes (such as close-ups of characters or micro-expressions against static backgrounds) place far lower demands on per-step denoising precision than high-dynamic scenes, making it easier to achieve good results at low step counts.