v1015

•This model is a style-based model based on sd xl1.0 with a pseudo-realistic style

•Recommended dimensions:1024*1024 or larger, 6*1280, 768*1360, 1024*1280, etc.

•Recommended steps: starting from 25

•Recommended sampling: local[restart][dpm++ 3m sde karras]

•lib online[Euler a]

•Antonym recommendation: lowres, (Style-Japan: 1.2), Paintings, sketches, lowres,

•cfg scale: starting from 7

•Updated on October 18, 2023, I will continue to iterate and improve

•Enlargement recommendation: Use image generation for enlargement processing, maintain consistency in sampling and image output, use 1.5-2 times the steps of the output, and redraw with a range of 0.1-0.2

•Note: Feedback from group members finds it difficult to produce full-body images, there is a slight bias towards upper body training images. In such cases, please use (thighs, bare feet, high heels, etc.) to add some lower body vocabulary, or directly use((wide shot, full shot, lower body, standing, stylish pose)), full-body AD retouching recommendation:[face_yolov8n_v2.pt], for large-scale image processing, it is recommended to enable AD. Most of my high-definition images are processed by Ultimate SD upscale enlargement, and the enlargement algorithm is[4x-UltraSharp]

----------------------------------------------------------------------------------------------------------