FLUX: just use 'natalie portman'

(everything below is sd 1.5)

SD 1.5:

Minimal tags to start / trigger word: n@tp0rt

clip skip 1 / cfg 4-8 / steps 15-75 / ADetailer inpaint denoise 0.1 / 512x512, 768x512, 512x768

DO NOT use WOMAN or 1GIRL in your prompt - the model is trained with the 'woman' class, so adding another 'woman' will be like adding 'woman, woman' - which can cause weirdness.

Example:

Positive prompt:

<lyco:n@tp0rt:1> a photo of n@tp0rt standing at a bus stop in a long coat

ADetailer positive prompt: (same as main but remove non-face-related info):

<lyco:n@tp0rt:1> a photo of n@tp0rt

Negative Prompt:

try with an empty negative prompt first - maybe add 'young', 'baby face', or 'child like' if it's turning out images that look too young.

Tested with:

realisticVisionV40_v40VAE

edgeOfRealism_eorV20Fp16BakedVAE

icbinpICantBelieveIts_afterburn

pirsusEpicRealism_v20HyperDetailed

aZovyaPhotoreal_v2

analogMadness_v30

analogDiffusion_10Safetensors

absolutereality_v16

ScenicOverlookOfRealism

96 tagged images sized 512x512, 768x512, 512x768. trained for 5760 steps

(permission for commercial use set to none to avoid copyright issues falling back to me - research purposes only, blah blah blah - I actually don't care what ppl do but don't go saying you had my permission to do it...).