在训练过程中，我使用了"ootachi"和"holding ootachi"等关键词来呈现拔刀场景的效果，并通过"ursula mask"关键词来实现角色佩戴面具的状态，但是对于多个武器的携带、握持和面具形态学习不是很好，可以实验性使用。目前模型对人物本身的刻画能力尚可，使用bodysuit来唤起角色本身的服装。但仍存在一些问题。有时模型会将角色与明日方舟中的史尔特尔混淆，导致头部可能会长出不必要的角。示例图像生成时采用了1.0的权重参数，其他权重效果有待进一步测试。如果您希望生成单人图像，建议添加"solo"标签以避免出现多个角色。如果不希望看到多个奇怪的剑柄和刀鞘，可以尝试在negative prompt中添加multiple weapons。敬请享受厄休拉的lora！<h2 id="-hfnng9rjq"></h2>During training, I used keywords like "ootachi" and "holding ootachi" to render sword-drawing scenes, and implemented "ursula mask" to depict the character wearing a mask. However, the model doesn't learn very well when it comes to carrying/holding multiple weapons and mask forms, so these features should be considered experimental.Currently, the model performs reasonably well in capturing the character's features - use "bodysuit" to evoke the character's original outfit. Though some issues remain. Occasionally, the model confuses the character with Surtr from Arknights, resulting in unwanted horns appearing on the head. The example images were generated using a weight of 1.0, with other weight parameters yet to be thoroughly tested.If you want to generate images with just this character, it's recommended to add the "solo" tag to prevent multiple characters from appearing.If you don't want to see multiple strange sword handles and sheaths, try adding "multiple weapons" to your negative prompt.Enjoy Ursula's LoRA!

Files with SHA256: ca533883c50ba6d10030dc7b24f414948677bf3ee943d5c4555c3ad983ece894