In most cases no other prompt words are required
1. Works best with a weight of 0.5
2. Does not affect other lora
3. Perform well on 2D models
If the picture quality is bad:
1. Add several similar prompt words: shirt, blue eyes, bow, braid, sleeveless, sleeveless shirt, smile, long hair, black bow, skirt, bowtie, hair ornament, one side up, white shirt, bangs, bare shoulders, collared shirt, pink hair, frills, medium breasts, side braid, black bowtie, hair bow
2. Modify the weight to 0.6
Schematic images are generated using the same set of prompts
Use mmd to train
~~o(≧ω≦)o
v2.1:set the block weights based on v2.0