A new merge, this time by using Perturbed Attention Guidance. (https://github.com/cvlab-kaist/Perturbed-Attention-Guidance). I have added a DMD2 checkpoint with very little weight to get amazing results but this is NT a DMD2 model.

Prompt: RAW portrait Photo of a woman, semi-side view, looking at viewer, wearing a maroon sweater, dark Paris sidewalk background, night, bokeh
Steps: 30, Sampler: DPM2 a, Schedule type: SGM Uniform, CFG scale: 5, Model: ZEP11, VAE hash: a30a98f479, VAE: sharpnessAndContrast_v10.safetensors, Version: v1.10.1

The model does not render few art styles like Zep9 but has a new look especially for people related prompts. This model can do NSFW and need users to use prompts and negative prompts to get SFW content. Enjoy!