Here I am sharing my currently go-to model merge for creating realistic NSFW images. I will provide the model links as well as the merge ratios and some generation settings I use. I use Forge so if you use another interface results may vary.
My examples only use the model. No loras, embeddings or anything else is used. Also no ADetailer or Hires fix is applied. Just the prompt and the model. I suck at prompting so I am sure you can get much better results than I am showing below.
If you have any suggestions for other merges, drop them below!
Currently I use a 50/50 mix of NatViS: Natural Vision and CognitiveRiver v2:
https://civitai.com/models/617652/natvis-natural-vision
https://civitai.com/models/563148?modelVersionId=683952
My preferred parameters are:
CFG: 4 - 6
Sampling method: Euler A AYS or DPM++ 2M SDE Heun Exponential
Sampling steps: 30 - 50
