Kou Tsubame LoRA
Did a bit of a test with this one, the dataset of this lora includes a significant amount of chibi or otherwise very basic or rather shitty art. The result, it seems to me, is that some parts of it are significantly less overbaked than they would have been without the inclusion of those images. The consequence is that generations will end up being more flat and simple looking compared to what you may be accustomed to with the model you're using. Some things counteract this, like adding a high weight "3D" into your prompt, using controlnet, etc. (or you can try putting the weight down, I tend to use all my loras at weight 1).
Due to the dataset, her jacket shows up less often than it doesn't when it's not prompted for, but having jacket in negatives and/or bare arms in the prompt will make that a certainty. Her shirt is "see-through shirt".
I found that trying to add "lightning bolt symbol" to get the lighting bolt in her eyes and bikini top actually just resulted in random lightning bolts in the environment or on her clothes, not including this tag seems to generate the lightning bolt on her bikini top just fine, while the eyes just never end up having them. I may or may not be able to do anything about that at the moment.
As always, examples are included in the folder
*ahem*
TOMBOY ROBOT SEX
https://mega.nz/folder/dfBQDDBJ#3RLMrU3gZmO6uj167o-YZg/file/5e4QjTrS 