A lora trained on images from the anime Kaiba, can be used to make cute characters and abstract backgrounds

There are two versions

V1 - My first attempt, makes cute characters in the style, not as great for backgrounds. This version is better for mixing loras.

V2 - Larger/cleaned database, very good at backgrounds but may needs a lower weight to get characters when mixing loras.

I tested on multiple checkpoints/loras, it plays nicely with almost every lora I've tried and most 2d oriented checkpoints. It's hit or miss with a mixed one like revanimated and mostly bad with anythingink. I got the best results with hellowflatcute2d. I am new to making loras and included the training data in case anyone wants to improve on it. Although there are tags for the main characters from the show this lora isn't very good at making them.

You should be able to make better images than what I've shown, I can never replicate the same example quality of loras I download :/

suggested tags for bg elements(v1 only accepts bubbles and planet)

• memoryroe (yellow/orange specks)

• planet

• smoke

• buildings

• patterns

• bubbles