For best outputs and usage:
Sampler: DPM++ 2M Karras
Steps: 25, cfg 7, CLIP skip 2.
VAE: sdxl_vae.safetensors.
============
This model has 2 main looks, cartoon and semi-realism. You can switch between them as follows:
For semi-realism:
Pos prompt: add this at the end the prompt "score_9, score_8, score_8_up, (score_7_up:0.5), (score_7:0.5), smooth_lighting, (score_6_up:0.5),
Neg prompt: this one "makeup, zPDXL-neg , score_3, score_4, score_5," plus what you don't want to see, being the "zPDXL" embedding included in my semi-realistic example runs.
============
For cartoon style:
Pos prompt: add this at the end of your prompt "score_9, score_8, score_8_up, score_7_up, score_7, sharp lines".
Neg prompt: this one "simple_background, score_6, score_5, score_4, score_3, score_2, score_1, censored, patreon username, patreon logo, patreon web address, signature, artist name, pony_horn, messy color, deformed fingers, bad, graphite (medium), monochrome, sketch,".
============
Despite it being a Pony Diffusion merge, the model doesn't listen as well as other Pony merges, therefore you do may need to add weights or try to explain it what you want in a different way.
The recommended sizes are 1024 x 1024 / 768 x 1024