This is an Lora of an Chinese artist Himitsu's style,it is trained based on Pony Diffusion and it can only work with it.

Recommended parameters:

Sampler:Eular a

Steps:25-30

CFG scale:4-7

LORA WEIGHT:0.4-0.6(very important)

Prompt:

score 9,score 8 up,score 7 up,score 6 up,source anime,2d,himitsu,{describe what you want}......

Negative prompt:

worst quality,low quality,worst detail,low detail,source_furry,source_pony,

For those who are not proficient,I highly recommend you use the above parameters,they may not be the best,but they are the most stable.ESPECIALLY the Lora weight,if it's too high the generated image will become distorted and strange,there may be different suitable parameters in different situations,as a personal suggestion,I often use 0.6.If you have better suggestions,you can put them in the comment section.

This is the first version of this model,overall,it can do well in imitating this particular art style,but in the aspects of the roles,I noticed some overfitting phenomena,maybe you should provide more detailed description of the role(such as the color of pupils),rather than simply a name of a role.Maybe I will fix these issue in upcoming version.

Search AI Models & Addons

very few results