Shampoo - Ranma 1/2

A LORA for our favorite Chinese amazon. Who I consider the best of the 3 fiancees considering her interests. Unpopular opinion: Akane sucks and Ukyo is bland! My favorite Waifus in the series are in no particular order: Kasumi, Shampoo, Natsume "Tendo", Ranma-chan and maybe Hinako.

V2:

The biggest change is the tagging for outfits.

Had good results with: <lora:ShanpuuRanmaV2:0.7> shanpuuranma, large_breasts, <One of her outfits.>

Available outfits are:

• Short_Dress_Pink_SMP, China_Dress: Her pink minidress, "surprisingly" most of the images i found of it are torn so it is best to use is with China_Dress otherwise it might deviate.

• china_dress_red_SMP: Her red china dress, I don't remember exactly where it is from.

• China_Dress_Golden_SMPOVA: Her golden dress from the reversible jewel ova, i think. It might be an actual weeding dress but i am not sure.

• Chinese_Clothes_Waitress_SMP, Apron_Frilled_SMP: Her Nekohanten outfit, I tagged her apron separately.

• Chinese_Clothes_Waitress_SMP: same as above but without her apron.

• Chinese_Clothes_Floral_Print_SMP: Another one of her Nekohanten outfits, this may require to add pants as it sometimes comes as a dress version.

• Chinese_amazon_warrior_Outfit_SMP: Her chinese amazon clothes with bustier armor, it is a bit iffy and might also require the tag pants.

• Bikini_Apron_SMP, Apron_Frilled_SMP: Her Nihao my concubine beach outfit.

Surprisingly two of the outfits i expected where not here but it probably was the Mandela effect making me think they were better that they really were. Those are her second outfit of nihao my concubine as most of the girls dresses are quite nice and the one of the Christmas special which most of them were crappy actually.

Anyway, maybe I shouldn't have retrained this LORA as when i checked the old one, the parameters weren't that bad. Only thing missing from that one was the tagging of some of her iconic outfits. Well that has been solved. Regardless V1 should be a much more flexible one. Any crappiness in the sample images is because at that point in time I still thought that sample images were just that, samples which showed the bare minimum you could get from a LORA. Alas people are a bit superficial and i get much better traffic if instead of the bare minimum which a beginner is expected to get, i input the bare minimum normal ones with my full litany of negatives and positives. This feels quite dishonest but if it is what people want then that's fine i guess. So fuck the noobs i guess...

Regardless! I have added the full dataset with the tag summary, so go read it so you can experiment and know what you can make with this. For example her hair_bell is in there as well as a couple of weeding dresses and bathing suits. Nothing iconic or with enough images as to add an extra trigger but still stuff you can make with a bit of effort.

V1:

The Lora seems to behave from .3 to .8 in between it will go from generic to more retro. For my tests i settled on .7 for the slightly more retro style.

If you don't add purple hair her hair color seems to gravitate to blue.

Had good Results with: <lora:ShanpuuRanma:.7>, ShanpuuRanma, big_breasts, hair_bell, china_dress,

Interesting tags are:

• china_dress

• chinese_clothes

• dress

• couple other outfits, just check the file

I added an XYZ plot so you can see the LORA weight effects.

I'll add the tag summary inside the image dataset.