Sampling steps

Sampling steps are the number of passes a diffusion model makes to turn noise into an image or video. Fewer steps mean faster results, usually with some loss of detail or variety.

  • Using models
  • Updated

A diffusion model starts from noise and cleans it up a little at a time. Base models are typically run with dozens of steps. MOSS-Image defaults to 50, and Anima suggests 30 to 50 for its regular version.

Distilled versions compress that work into far fewer steps. Z-Image Turbo and Qwen-Image-2.1-Turbo need 8, Anima's Turbo version 8 to 12, and MiniMax H3-Image and Pika H3 Flex 6. Those versions are the practical choice on a single consumer GPU, but they usually give less varied results and fewer controls than the base models. When you compare settings, change the step count first, because it multiplies the time of every image.

Related terms

  • Diffusion model A diffusion model generates images or video by starting from random noise and removing the noise step by step until a picture emerges.
  • Classifier-free guidance (CFG) Classifier-free guidance, usually called CFG, is a setting in diffusion models that controls how strictly the result follows your prompt, at the price of extra computation.
  • Distillation Distillation trains a smaller student model to imitate a larger teacher model, keeping much of its ability at a fraction of the size.

Keep reading

All terms