An attempt to create a single lora that would allow for the generation of both Roko-Chan and Yudko-Chan. Whether alone or in the same image. In general I think that the solo lora work better for the individual characters. However, this combo lora can still handle them both in isolation or together. Though there's a significant difference between the diffusion models. The Anima , Z-Image Turbo , and Krea 2 lora also share some issues with Roko-Chan's tail often coming out nonsensical. Overly coiled, no real end, etc. Hopefully I'll have that ironed out over future versions. The Krea 2 lora is the best at avoiding the tail problem but it's far from perfect.

Krea 2: The best option at the moment. It can generally handle both Roko-Chan and Yudko-Chan fairly well.

Anima: A solid second place.

Z-Image Turbo: Often struggles when trying to use both characters in a single image. In particular with one of them having robot legs and the other having the snake base instead of legs. However, it handles them solo fairly well. With Roko-Chan still being a 'bit' unpredictable as far as her shirt and tail coming out correctly.

A retrain is planned for at some point which will leverage very careful inclusion of synthetic data to help deal with the tail issue.