I made a dumb decision to split XNH into a collection, so these versions will have to be separate models until Civit hopefully allows you to convert models into versions.
This is a merge that I've made because I couldn't choose a favorite model, which has led to not being able to choose a favorite of my own models (-_0)
It's pretty good, so I'm sharing it :)
It should be good at following prompts. I've no idea how it works with LORAs or artist styles, I don't use them.
Avoid the "Bad" models. I should delete them, but I'll keep them for whoever's interested.
More info in "About this model" section. Whichever preview image has the most reactions will be the cover image...will probably end in the rich getting richer :p
5-7 cfg
20-30 steps
Probably any normal SDXL resolution, I use 832x1216
Positive: masterpiece, absurdres,
Negative: (bad anatomy:1.5), signature, watermark, (worst quality, bad quality:1.2),
Negative-2: sketch, worst quality, bad quality (slightly different style, may make better images/follow prompts better than negative 1.
Other styles:
Photorealistic // Flat color // Anime screenshot // 3d // Western comic /(style/) // Furry
Description
A merge containing LORAs and much confusion. A brand-new style, but sometimes follows prompts worse or better than 005. Use cfg 5 to avoid potential anatomy issues and glitches.
This model: https://files.catbox.moe/pgipsf.png
Stage 3 - Stage 2 + cunny & Flat
Stage blahblah (Stage 2 006): https://files.catbox.moe/b0oc1r.png
Epsidaddy should be the Abmayo lora + 005.
I didn't save the workflows of the intermediary models before I deleted them :/
LORAs used (all that matters anyway):
Abmayo.safetensors by Waikitty
Theobrobine Illustrious.safetensors by ShadowxArt
cunnyfunkystyle2.safetensors by ofake
illustrious_flat_color_v2.safetensors by Motimalu