For best outputs and usage:

Sampler: DPM++ 2M Karras.

Steps: 35, cfg 5.5 ~ 7, CLIP skip 2.

VAE: sdxl_vae.safetensors

Pos: start it with "score_9, score_8_up, score_7_up, source_real, rating_explicit, detailed, realistic photograph, highres photo of" and end it with "realistic, volumetric lighting, intricate details, epiCPhoto, ", being the embedding "epiCPhoto" and the LoRa "add-detail-xl" included in m example runs.

Neg: "score_3, score_2, score_1, source_anime, source_cartoon, disembodied penises, mosaic censoring, bar censoring, censored, text, signature, logo, artist_name, username, crop, frame, (simple background:1.6), worst quality, low quality, normal quality, messy drawing, amateur drawing, lowres, bad anatomy, bad hands, source filmmaker, 3d, showing teeth, censor, bar censor, mosaic old, granny, wrinkles, censorship, fat, BeyondSDXLv3, epiCNegative" being the embeddings "BeyondSDXLv3" and "epiCNegative" included in my example runs.

Checkpoint Merge.