This time I made it with anime illustrations because I couldn't generate what was I looking for with the existing LoRAs. It's not perfect (as you can see on the model main image hands) but can make some really good images.
When generating NSFW you should use "uncensored" tag because the dataset its almost fully SFW.
Update 1.2 for WAI-NOOB-SHUFFLE VPRED 03:
Trained on the model 03 new version.
Update 1.1 for WAI-NOOB-SHUFFLE VPRED 2.0:
Added a lot of NSFW to dataset.
Should be a lot better in general.
For some reason the images don't look the same in my local ComfyUI compared to the online generator but it uses the same settings.
Recommendations for WAI-NOOB-SHUFFLE VPRED 2.0/3.0:
Scheduler: ddim_uniform
Cfg: 3-5
Recommendations for NoobAI:
Sampler: Euler
Scheduler: Normal/SGM Uniform
Take it as a first experimental version, will try to improve it.
Can also add "jaggy lines" to negative prompt and/or "clean lines" to positive prompt for a clean lineart because many of the illustrations on the dataset look handdrawn.