π LoRA is trained purely on anime screencaps
π Activation tags are in the right panel
π Note:
Check preview images for detailed positives & negatives combination
Includes most of her outfits from season 1
Too many print outfits and not enough tags to make each unique, but most of them generate more or less stable as long as you use correct negatives.
The only unstable print is: MomoWhiteSweater, the picture on that sweater is very random
πΈ Always use ADetailer to get detailed eyes
π Technical notes:
LoRA is compatible with various styles - Lower the weight & remove "anime screencap" from the prompt, if you are getting too strong anime style
Hi. res fix used on all sample images:
Upscaler: R-ESRGAN 4x+ Anime6B
Steps: 10-15
Denoising: 0.3
π¬Leave a like, comment, review but first of all, ENJOY ;)