It is trained on anime images that are close to 2:1 wide, with bold composition, dynamic, and oblique angles.

Due to some problems in diffusers training lora, the current effect is only to test whether the weights can be loaded correctly