Playing with inference steps
wasting time with style
As said on Basics of SD, one of the most important parameters that can affect the image quality is the amount of time the AI is allowed to spend on an image.
So let's play around with generation steps a bit, and see what happens... (this parameter is actually called "cycles" on @ai621bot)
Let's start with these parameters:
Prompt: canine anthro furry wolf, dark white and grey/brown fluffy fur, spreading legs showing bulge on black and cyan boxer_briefs. showing armpits. happy smile, closed eyes, tongue out. solo. simple background. male with navel showing. on_bed
Negative: penis explict know genitals
Seed: 9499787, Sampler: euler_a, Inference Steps: 20, Guidance Scale: 4, Model: yiffy-e18, Negative Prompt: penis explict know genitals
And we get, a wolf!
Play with the slider below to see how much training steps affect generation quality.