
Trained to resemble official artstyle.
I've not done extensive testing but some issues (varies per model and strength):
-
her dress shirt seems to have baked itself in - sometimes the collar, sleeves or center frills will appear unprompted
-
the white spots on her hair might appear in too large quantities
The DAdapt version seems slightly less accurate to the style/more generic and I think it avoids these issues better, so use that one if you run into them too often and lowering strength doesn't help.
Miyuki-2000 is old, but included it as it is the one I used for my pixiv pictures.
When prompting for "crazy eyes", putting "red eyes" in negatives might help, but unsure.
Prompts used for her dress shirt
-
collared dress, white dress, dress, center frills, belt, long sleeves
Prompts used for cardigan/school uniform
-
yellow cardigan, neckerchief, serafuku, school uniform, pleated skirt, buttons, skirt, sailor collar, miniskirt
You probably don't need all of them - the first tag should be enough.
Prompts for the red scenery and sky
-
red theme, outdoors, no humans, building, scenery, chain-link fence, surreal, monochrome, rooftop, cloudy sky, red sky
Same as above, "red theme", "monochrome" and "surreal" should be enough, include sky in either positive or negative as required.
Feel free to contact me if you want the dataset or for any other purpose.
描述:
训练词语: Miyuki
名称: Miyukifinal-000010.safetensors
大小 (KB): 147578
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success