Use stable diffusion base mode to replace chilloutmix for training, and use more images.









Trained on more images, and use sd1.5 as basemodel other than chilloutmix, works much better for anime models.
!!warning: stable diffusion are not good at hands, try multiple time.
Try and share your results, let me know how it works
================
V1.0 is an early version, may generate wired hands/penis, and has low successful rate (~20%). But when it works, it works great, with very good results. So I think it still worth to share.
Always try multiple time to get best.
(多试几次,如果有效,图片会非常不错)
===============================================================
licking my ????, <hair style, eyes, any other styles, background you like>
Recommend to always add some description about the hair and face, otherwise, the model will likely to generate 2 penis~
v1.2 trained on sd1.5 base model, can work across many different models.
v1.0-EA Trained and tested in model: https://civitai.com/models/6424/chilloutmix Not sure if other model works.
0.7~1
licking my ????, white woman, red head, clear eyes
Negative prompt: two penies, illustration, 3d, sepia, painting, cartoons, sketch, (worst quality:2), (low quality:2), (normal quality:2), lowres, bad ???????, bad hands, normal quality, ((monochrome)), ((grayscale:1.2)), futanari, full-package_futanari, newhalf, collapsed eyeshadow, multiple eyebrows, vaginas in breasts,holes on ???????, fleckles, stretched ???????, ??????? on ????????, analog, analogphoto, signatre, logo,2 faces
Steps: 30, Sampler: Euler a, CFG scale: 7, Seed: 331484549, Size: 512x768, Model hash: fc2511737a, Model: chilloutmix_NiPrunedFp32Fix
Use stable diffusion base mode to replace chilloutmix for training, and use more images.
