thanks to GOOGLE translate,luvs!

This model is a mixture of several popular models and contains my own interpretation of the aesthetic style. I aimed to create an image style that was close to reality but without the unsightly aspects of real-life portraits. It also had to be able to recognize different portraits, rather than like a cartoon where all the characters are the same face. After testing, it is compatible with most Asian portrait lora models.

I must admit it's more suitable for nsfw content, so get your papers ready and have fun!

Recommended settings

Tip words: Any common ones you can think of will do

Generate settings: 2m or sde sampling (karras)

If the generated image is not clear, reduce the cfg appropriately, but 7 or more will be better.

Keep the image size as large as possible

use a magnified model if necessary

if you like the model，follow YORUSHIKA,they are a good band.