I would highly recommended using both of trigger words, u usually will get better results.

Also, couple of recommendations:

• Use lawlassYiffymix20 for results as showed on the page

• recommended clip skip 2

• set around 0.9 scale (I was using exactly 0.9)

• the model trained on 512x512 images

If you dont get good results, write an feedback with your generations. I'll try to fix this.