I's my third attempt to make a proper lora but i guess it's time to move on. My goal was from 3d pictures take hairstyle and facial features but at the same time get rid of 3d artifacts at least partially. I trained it with batch size 3 and picked 1300 step. More steps and it's starts to destroy background. I planned to use just 映雪 as character key but it's take longer to train and while it's training longer it have more opportunity to mess up everything, especially backgrounds. Usage notes: Strength less then 1 make character loose it's features but strength 1 often came out slightly broken. Main tags: 映雪, vam, white hair Other tags to play: 3d, beautiful rendering, white pantyhose, dark background, light beam, see-through, blue light, portrait, upper body, silver bra, holding dress

VAM character 映雪

Trained and tested only on vanimix_v15 checkpoint

Files with SHA256: 94665b600b9dd6968922df5e92a4b90376c622b98ca7f817edee0853c34bbaa7