PiroxiumDiffusion版本v1.7 (ID: 125085)

General use model trained using offset noise, uses default stable diffusion VAE.

I recommend playing with the strength of the trigger word eg: (PiroxiumDiffusionStyle:0.7)

描述:

Recommended Settings:

Clip Skip: 1

CFG scale: 9-11 (This model seems to have a mind of its own at lower CFG scale)

Sampler: DPM++ 2M Karras

Final Advice:

Place "PiroxiumDiffusion style" at the beginning of prompts for stronger model influence

This model does outdoors best, indoors can be a struggle sometimes. Portraits work best. The model was trained with captions using wd14 tagger.

The preview images are some of the best images I was able to create with this model. Most images will probably not look as good.

训练词语: PiroxiumDiffusion style