This model should control the difference in size between a female and male character.Main prompts:size difference, larger female, smaller male,tall female: a woman is slightly taller than a manAmazon: a woman is significantly taller than a man (approximately at the level of the stomach-chest)Amazon, half-size male: an intermediate option.half-size male: a woman is twice as tall as a manthird-size male, half-size male, an intermediate optionthird-size male, a woman is three times taller than a manthird-size malle, quarter-size male, an intermediate optionquarter-size male, a woman is four times taller than a manquarter-size male, mini boy, an intermediate optionmini boy, the difference in size is more than fourfold.Various auxiliary hints have also been added: girl on top, on back, on stomach, lying on person, boy on top, on back, on stomach, lying on person, on side, on back, lying on person, sitting on person, on stomach, sitting on face, hand on another's head, carrying, head between breasts, breast smother, between thighs, girl sandwich, sandwiched,In the version for Anima, the dataset was expanded: the amazon tag was replaced with mini giantess, and the giant female tag was also added, which is recommended to be used together with the tags third-size male, quarter-size male, and miniboy. The rest remained the same as in the previous versions.

Size difference (larger female)

Well, here’s the third version of my LoRA about the size difference in favor of girls, this time based on Anima. Yes, Anima without any LoRA already handles drawing big girls quite well, but still not well enough in my opinion. In this version the dataset was expanded, the tag “amazone” was replaced with “mini giantess”, and the tag “giant female” was also added — it’s recommended to use it together with the tags “third-size male”, “quarter-size male”, and “miniboy”. Everything else remained the same as in the previous versions. I can’t say the result is ideal, but in my opinion it’s acceptable. I don’t know whether I’ll continue refining the LoRA or just leave it as it is, simply because everything comes down to time — both mine and my computer’s — as well as further expanding the dataset (after scouring a bunch of sites I barely managed to scrape together 720 arts, and finding new ones is getting harder every time).

Files with SHA256: cbd85064fbdb7989eb57ce6c16ccbee4f75f0718aaf1c1f30d8f4c65af66f415