Tang San - Douluo Continent

1. Notes for Image Generation:

①: Training material 512*896, to achieve the same effect, you must use the base model I used for training to generate the image! And be sure to use negative prompt words!

②: Training base model: National animation goddess

Sampling: DPM++ 2M Karras

③: For this Lora weight, refer to the cover image

④: ※For a half-length portrait, you can skip high-resolution restoration for a decent effect. If you want to add details by turning on high-resolution restoration, the model to use is SwinIR_4x, with a redraw range of about 0.3-0.4

If the generated image quality is not good, feel free to leave a comment in the section below, I will help resolve it when I see it