画質のよい絵を出す目的ではなく、アニメっぽい絵を再現することを目的としています。
一部の衣装はメインタグ(yuukiyuuna:1.0)の比率を下げないと再現できないことに注意。
LoRAの学習はIllustrious-XLですが、サンプルの画像出力はWAI-NSFW-illustrious-SDXL(v8.0)です。
Illustrious-XL:https://civarchive.com/models/795765
WAI-NSFW-illustrious-SDXL:https://civarchive.com/models/827184/
The image quality is low because the purpose is to reproduce an anime-like image.
Please note that some costumes cannot be reproduced unless the ratio of the main tag (yuukiyuuna:1.0) is lowered.
LoRA training is done with Illustrious-XL, but sample image output is done with WAI-NSFW-illustrious-SDXL(v8.0).
prompt
<lora:yuukiyuuna-ILXL-v2:1.0>,(yuukiyuuna:1.0),(red_hair,red_eyes,ponytail,hair_ornament:1.0),(white_school_uniform,pleated_skirt,short_sleeves,red_bowtie:1.0),
<lora:yuukiyuuna-ILXL-v2:1.0>,(yuukiyuuna:1.0),(red_hair,red_eyes,ponytail,hair_ornament:1.0),(gray_school_uniform,gray_cardigan,white_sundress,long_sleeves,red_bowtie:1.0),
<lora:yuukiyuuna-ILXL-v2:1.0>,(yuukiyuuna:1.0),(red_hair,red_eyes,ponytail,hair_ornament:1.0),(gym_uniform:1.0),
<lora:yuukiyuuna-ILXL-v2:1.0>,(yuukiyuuna:1.0),(red_hair,red_eyes,ponytail,hair_ornament:1.0),(hooded_hoodie,shorts,printed_shirt:1.0),
<lora:yuukiyuuna-ILXL-v2:1.0>,(yuukiyuuna:1.0),(red_hair,red_eyes,ponytail,hair_ornament:1.0),(bikini:1.0),
<lora:yuukiyuuna-ILXL-v2:1.0>,(yuukiyuuna:1.0),(red_hair,red_eyes,ponytail,hair_ornament:1.0),(pink_kimono:1.0),
<lora:yuukiyuuna-ILXL-v2:1.0>,(yuukiyuuna:0.8),(red_hair,red_eyes,ponytail,long_hair:1.0),(white_japanese_traditional_clothing,head_ornament,:1.0),
<lora:yuukiyuuna-ILXL-v2:1.0>,(yuukiyuuna:1.0),(red_hair,red_eyes:1.0),(pajamas:1.0),
<lora:yuukiyuuna-ILXL-v2:1.0>,(yuukiyuuna:1.0),(red_hair,red_eyes:1.0),(anniversary_dress,hair_crown,pendant:1.0),
<lora:yuukiyuuna-ILXL-v2:1.0>,(yuukiyuuna:1.0),(pink_hair,ponytail,long_hair,red_eyes:1.0),(mankai:1.0),
<lora:yuukiyuuna-ILXL-v2:1.0>,(yuukiyuuna:1.0),(pink_hair,ponytail,long_hair,red_eyes:1.0),(yuusya:1.0),
<lora:yuukiyuuna-ILXL-v2:1.0>,(yuukiyuuna:0.4),(pink_hair,ponytail:0.8),(dai_mankai,very_long_hair,heterochromia:1.2),
Description
FAQ
Comments (15)
ありがとうございます🙏🙏🙏
助かりました😭
thanks....illust model for yuuyu...i need more from this series hehe
I plan to make Sonoko next. If there is a character you want me to make in the YuYuYu series, I might make them a priority.
@NARARILAI sonoko...yes please....lily yuuna also good or togou
@GlennRadars I already have a data set for Togou, so I think I can make her right away. Lily_Yuuna is an unexpected character. I don't know if I can make it well because there are only a few materials, but I'll make it.
@NARARILAI nice...thanks
also how much image you need to train a model/character? for SD, PONY and ILLUSTRIOUS? and how much VRAM you have?
@GlennRadars I've heard that 20 images is optimal.
But you can do a little with just one image.
My PC's VRAM is 16GB. I think the VRAM usage with SD was about 8GB, but I've never seen the VRAM usage with SDXL.If I haven't forgotten, I'll take a look at it today when I do machine learning.
@NARARILAI ahhhh SD take that much vram... u_u
@GlennRadars I can't remember exactly, so I'm not sure. It might have been 6GB.
ILLUSTRIOUS consumed 9.4GB.
@NARARILAI 9GB...............~~
@GlennRadars Since I had the chance, I chose Lily first,Since there were few samples, the eyes were strange and I couldn't reproduce Yuusya clothes, so I tried learning it about 10 times, but this was the limit.
@NARARILAI well yes the eyes looks strange and yuusha clothe too but its ok maybe it can be improve time by time and now i can generate lily....huehehehe
she short af
is there any reason why i always get different dress for annivesary dress?
I don't quite understand the question. Do you mean that the clothes of the anniversary dress change every time the image is generated? This is because the anniversary dress used for learning has only two learning images, so the degree of learning is low.
ohh only two...thats the reason........well ok then....well i just add post one of anniversary dress that i get...well its beautiful tho
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.