This is a hybrid model that I usually use myself, mixing several common 2.5D models. The model has good expressive power on themes such as realism, fantasy, and NSFW.
这是我平时自己用的一个混合模型，混合了几个常见的2.5D模型。模型对写实，幻想，NSFW等主题都有较好的表现力。
IMPORTANT MATTERS(重要事项)
To ensure compatibility with checkpoint, I did not incorporate any Lora.
为保证checkpoint的兼容性，我没有融入任何lora。
Suggest enabling high-definition repair! HD repair: 2x, redraw range: 0.4-0.5.
建议开启高清修复! 高清修复: 2倍, 重绘幅度:0.4-0.5。
The sampling method is recommended for the DPM++series, with steps of 20-30 and CFG of 7-8.
采样方法建议 DPM++系列 , 步数20-30, CFG:7-8。
建议分辨率:512X768
Recommended resolution: 512X768
Description
v2.3
1.Starting from version 2.1, the model has been changed to. safesensors format. Version 2.3 is my most commonly used version, and the 2.5D effect is the best.
2. It can also be used as a training base for 2D or 2.5D Lora, with a very high degree of restoration to the training set style, but low generalization. Even if using other base models to generate images, it will still have a strong training set style. If you want to train your art style, you can consider using this model. If you don't need to train your art style, please use it with caution.