Sampling Steps Parameters
The number of denoising iterations during image generation. More steps generally produce higher quality but take longer to generate.
Sampling steps (or just “steps”) control how many times the model refines the image during generation.
How Steps Work
- Generation starts with random noise
- Each step removes some noise
- More steps = more refinement opportunities
- Eventually, quality plateaus
Recommended Step Counts
| Steps | Result | Use Case |
|---|---|---|
| 10-15 | Quick drafts | Previewing compositions |
| 20-30 | Good quality | General use |
| 30-50 | High quality | Final renders |
| 50+ | Diminishing returns | Usually unnecessary |
Sampler Matters
Different samplers need different step counts:
- Euler, DPM++ 2M: Good at 20-25 steps
- DPM++ SDE: Benefits from 30+ steps
- DDIM: Stable across step counts
- Ancestral samplers: Never fully converge
Quality vs Speed
More steps mean:
- Better detail refinement
- Smoother gradients
- Longer generation time
- Eventually no improvement
Tips
- Start with 20-25 steps for most work
- Use fewer steps when iterating on prompts
- Increase for final high-quality renders
- Watch for when quality stops improving
- Some models are optimized for fewer steps