An attempt to create a style.
You can vary the strength of the lora to get different results, optimally use 0.8-0.9.
Different cfg can also affect the result, I used 5-7 in the example images
Sometimes it can give a black and white result, you can add greyscale to the negative promt (but this may make the picture more colorful).
Description
FAQ
Comments (12)
Bro, thank you model, you are so good, fortunately.
You're welcome, if it's not too difficult, you can put it on the wall sometime so you can build your own generation with this model
@fakerrak I don't know English, do you mean to ask me to generate more pictures of this model？
@460672106622 Yes
@fakerrak It seems that the image generated using this style model is not the same as the image generated by the Su-Hyun HAN model
@fakerrak The images generated using the Su-Hyun HAN model are similar in style to the original manga
@fakerrak I don't know if my prompt is wrong？
@460672106622 it is, because this style model, it only gives certain facial features in the picture, if you want it to draw the characters like in manga, but you need to train each character in the models, I'm not very good at this
@fakerrak I use the Illustrious model to know the anime characters, and then use your style model, as if the anime characters will affect the style, what prompt words do I use instead to avoid this？
@fakerrak Thank you for answering.,It seems that you don't use anime characters.,Use prompt words that describe the appearance.,The style of painting is almost restored.
@460672106622 for the experiment, I generated some images of Wendy with the style of the models, which you can look at
@460672106622 if you want the style to influence the character more, increase the weight of the lora, for example, maybe try 1 or 1.2