Base on anime S2 and S3.
It seems with limited input I can make even better result...
better to be used in 0.5-0.7 I think.
More like the Mrs. yukinoshita one but I add one or two background this time. So the problem becomes there will always be something like window or stick or pillar in the picture. Try to use low weight or negative prompt to avoid it.
hires and blockweight is recommended as the example pictures use them. Nsfw is avaliable though I do not show it.
Description
FAQ
Comments (7)
What model are you using?
donkonsfwhard
Hi, I would like to point out a serious error. When I use the command: <lora:yuigahamaYuisMotherYahariOre_v10:1> I always get the pg from the side with the sweater. Even if I put Bikini in the Promts, nothing happens. Can you fix the Lora?
what is pg? That's how this lora is trained. Use lower weight or block to make it flexible. I make Lora like this because it's easy to lower the weight but hard to do the reverse thing.
大佬，用大佬的lora经常出现下面这个错误提示。
单张生成也有不低出现。使用其他lora多图生成也很少出现这个提示。显卡是笔记本2080S 8g。
modules.devices.NansException: A tensor with all NaNs was produced in VAE. This could be because there's not enough precision to represent the picture. Try adding --no-half-vae commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.
半精度问题和lora没关系吧
你是用動畫截圖來訓練lora的嗎
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.