Special day for me, so I made a more complicated merge. I would like to make a trained model at some point, but I am still learning how to make a decent dataset for one. This one take the excellent lighting on Pony FaeTality with a mix of anime tones.

Post your images and show me what the model can do. I am sure others have much better usability than me.

This model is combines these models with a simple recipe.

Pony 🦄FaeTality v1.1

Animagine XL V3.1

v1.1 removes Sora Ani2.5D Pony

v1.1 adds in Hassaku XL (Hentai)

General recommendations

Sampling method: I generally use Euler A / DPM++ 2M Karras (I have been experimenting with DPM++ 3M SDE Exponential)

Sampling steps: 30 ~ 50

Clip Skip: 2

CFG: 4-11

I like to use Pony PDXL Negative Embeddings otherwise use score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up.