
Trained on Animagine XL V3.1, 16 images.
This character was shown wearing only one type of dress. Sometimes, parts of that dress are generated despite not being included. She has some sort of horns/crown in the anime, I didn't bother much with that detail. Sometimes, that crown/horns are generated, despite not being included anywhere in the prompt, it's not a big deal.
In my testing, I found "lying on bed" or "sleeping" to be problematic, that might be because I used pretty closeup shots of Liala, which means AI will attempt to replicate the source material and prioritize covering the image with Liala's body.
I only used Animagine XV V3.1 to generate the images, it's my first time using the new SDXL stuff, so it will take some time to find new models that work.
I use Euler A. 34 steps. If you use PixAI or other currency sites, I recommend you go cheap first and then use more steps and stuff when you're sure you'll get the result you want.
8CFG
Thumbnail model: Animagine XV V3.1
https://pixai.art/model/1835881577390765602
描述:
PixAI trained LoRa
训练词语: Liala, blue eyes, long hair, wavy hair, light blonde hair, wool collar, white panties, see-through white cloth, wool wristbands, see-through white gloves, big breasts, small nipples covers,
名称: checkpoint-e100_s1300.safetensors
大小 (KB): 250055
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success