V2.0
child（幼年）
sylc,1girl,green hair,hood down, short hair,ahoge, pointy ears,red eyes,open jacket,long sleeves,shirt, black shorts
young（少年）
sylm,androgynous,white gloves,elf,ahoge,white hair,short hair,short hair with long locks,,pointy ears, tinted eyewear,cape,shirt,long sleeves,belt,black pants,
mature（成年）
sylm, 1girl, solo, pointy ears, elf, white hair, ahoge, white dress, cape, green capelet, red eyes,long hair, long sleeves,
————————————————————————————————————————
no tag
just 1girl
Description
FAQ
Comments (10)
动态显示发了个铜态点进去404 23333 有没有其他地方公开
没有没有，就这一个，就是那会c站犯病我发完刷新不出来，以为是对我标题起反应了，所以我删了改名重发的…
says "Couldn't find Lora with name sylphiette_v10"
I can't help you with this kind of problem...it may be a problem with the plug-in version
now i am really curious..
was using your lora after a very long time again to test a checkpoint. and..
i used "sylphiette" as prompt and it worked like a charm. no change in size or something that sometimes happens. but i got more curious if there were more users that made a sylphiette and so, i found you again and saw that the trigger is "1girl". so i used this prompt and there were absolute no difference.
my question would be, on what real prompt your lora was trained.
I use Google Translate, and I don't quite understand what you mean. Generally speaking, I will delete the character's name when training. This has two effects. If the A model knows "sylphiette", but the B model does not know "sylphiette", Deleting the name can allow lora to learn more characteristics of the character, and thus get better results in the B model. In addition, in some models, use "ganyu (genshin impact)", because "genshin impact" may cause large The model uses the characteristics of keqing or other genshin characters, so I only use 1girl, which will reduce the interference caused by the large model
For example, I recently trained the touhou and genshin characters. Although I suggest that you use names as trigger words, I deleted these tags during training. They are only used to help the large model recognize these characters, but they are not used in lora training. did not use
@kozue thanks that you took your time for a reply. i can fully understand you. what you mean is the same as if i use stable diffusion and type in "chibiusa"... results are just a huge mess but not what i wanted.
all right. still thanks for your reply, and p.s.: this sylphiette lora is absolutely awesome in its quality. mixing it with real life backgrounds was fun, because it gave sylphiette awesome details and awesome light effects.