Before going blindly into generating hundreds of pictures for an hypothetic LoRA, let's start with embeddings! (PS: It is not matching exactly the amazing character created by DarkFun since i had to group the description parts together instead of spreading them in the prompt, that's where a LoRA would do better.)

• TuliEmbed: the full description of the character

realistic face,smokey eye makeup,lidded eyes,female,sexy elite escort,italian and thai,feminine,brown hair,short wispy hair,gray eyes,realistic eyes,black lipstick,occult tattoos on arms and thighs,realistic skin,freckles,delicate chain choker necklace,slender physique,slim hips,small ass,small breasts

• TuliEmbedADface: the description limited to the face features for ADetailer

realistic face,smokey eye makeup,lidded eyes,female,sexy elite escort,italian and thai,feminine,brown hair,short wispy hair,gray eyes,realistic eyes,black lipstick,realistic skin,freckles

Warning: Those embeddings are HUGE. The first one is 68 tokens by itself (and the second one is 44, but that is less an issue in ADetailer prompt).

Make sure it is fully taken into account in a single "75 tokens slice" of the CLIP by using it at the beginning :D

HDR, high contrast, best quality, cinematic scene, solo, 1girl, TuliEmbed,
...

Once again, this issue would be fixed with a LoRA :D