This is Lora of Mitsuki Felicia from Magia Record.

The training image is 1672x957, so please generate it at a larger size, such as 1536x1536 or larger. Generating at smaller sizes like 1024x1024 will degrade the image quality.

Unlike yachiyo and tsuruno, there are no particular points to note or cautions.

Felicia's cheeky face can be reproduced with "smile, open mouth, screaming, happy, fang". Happy and fang are not essential, but try experimenting with the changes in expression!

According to GPT, Felicia has stronger features than tsuruno, so it was trained with a lower learning rate and conv_dim and conv_alpha than iroha, yachiyo, and tsuruno. In fact, slightly increasing the learning rate and conv compared to GTP's suggestion caused the face to stiffen. Therefore, slightly lowering the intensity results in a decrease in the accuracy of the costume reproduction. Because it was trained using only anime capture images, there are some scenes where the accuracy of reproduction is low in parts of the costume, such as the Kuroha costume and the school uniform skirt, because only a small portion of the image is visible.