I used stabilityai's stable-diffusion-xl-base-1.0 model with bunch of prompts and fine tuning in order to create these. Nothing else.
"Fine-tuning" has a colloquial definition and an AI definition. Do you mean you tweaked the prompts until they looked good, or do you mean that you fine-tuned the model itself on a handpicked dataset?
Also, these look like they were adapted from the base portraits, like the way an img2img model would do so. Did you just use prompts, or did you provide the original portraits for it to adjust?
In any case, these results look very clean, well done.