For best outputs and usage:

Sampler: DPM++ SDE.

Steps: 30, cfg 4, CLIP skip 2.

VAE: none required.

Pos prompt: start it with "score_9, score_8_up, score_7_up, ziprealism, photo, photorealistic", being the embedding "ziprealism" included in my example runs.

Neg prompt: "ziprealism_neg, CyberRealistic_Negative-neg", being both embeddings included in my example runs.

This model has the small disadvantage to take between 2-3 minutes to produce 3 outputs, due to the recommended sampler. It admits other samplers, but the recommended one will produce the best results visually wise.

Checkpoint Merge.