A few oddities. LoRA is not quite familiar to me. A little experimental. Initially, it was done on the basis of portraits, with specific glitches and tinting. However, sometimes interesting combinations can be obtained, depending on which specific model was used. I recommend strength from 1.5 and above. Sometimes, glitches do not appear, but only the tinting of the image and the characteristic pattern.

Files with SHA256: 80569bcf48feb974d65b353a9d1a045ebb0338cf8dd6297ba91563b8f8c4d155