Model was created using 482 curated images, 62 of those being full body and the rest cropped face photos. One thing I truly realized creating this was the importance, if you want full flexibility of the Lora, of cropping face photos. This was after running 3 revisions at a total of probably about 12,000 steps. The Lora worked but it was only after applying multiple creation techniques.

This version took about 7000 steps, and performs and looks vastly better. All sample images I'm posting at almost exclusively straight from txt2img with an application of Adetailer. Here are my recommended settings and tools I used in Forge.

Realism_Illustrious_By_Stable_Yogi 3.5

Realism_Illustrious_Positive_Embedding + Realism_Illustrious_Negative_Embedding

Realism_Lora_By_Stable_Yogi

Detail_Tweaker_Illustrious_By_Stable_Yogi (Pro version from his Patreon)

Euler A + Karras

26 Steps

CFG Scale 6

Width 1224 Height 856

Adetailer mediapipe_face_full - Inpaint mask blur 8

All of my prompts began with this or some descriptive variation

Positive

Stable_Yogis_Illustrious_Positives, realistic, photorealistic, realistic skin texture, detailed, indoors, dusk, light rays, indoors,

Negative

Stable_Yogis_Illustrious_Negatives-neg

text, 3d,2d, phone, rating:explicit

Being an Illustrious model it wants to be very horny...hence the last negative tag. Even with that, it still had some tendency to push through.