The model was created from 21 pictures from the video game Fire Emblem Engage, mostly in-game cinematics and artist's pictures. And was trained on AbyssOrangeMiX3 model from WarriorMama777
I used the Lora Training Colab from Hollowstrawberry and my own Dataset to train the model.
Here is a recommended sample prompt for Timerra (you may try other outfit, like dress, swimsuit ....) :
1girl, bike shorts, multicolored white pink crop top, navel, multicolored frilled_skirt, dark-skinned_female, jewelry, long_hair, ponytail, hair ornament flower hairband, multicolored beads bubbles, brown_eyes, symbol-shaped pupils star \(symbol\), (red_eyes:0.66), black_eyes ; <lora:timerra-v3-10:0.75>
Kaidras created a Lora for this character; I haven't tested it. I began to create the dataset two weeks ago, before he released it. This Lora is independent of his LoRa.
If you have any advice, please tell me.