该模型受到模型大佬LEOSAM的启发和交流，通过“差异训练法”训练得来。

不同的大模型会出现不同的结果，受限于当前技术，还是无法完全固定住人物的线条，所以图一是通过controlnet的canny得来，图二应该是使用anytingV5，大家也可以尝试使用其它模型。

Different large models may produce different results, and due to current technology, it is still not possible to completely fix the lines of characters. Therefore, Figure 1 is obtained through the Canny of Controlnet, and Figure 2 should use AnytingV5. You can also try using other models.

合作邮箱：[email protected]