Mostly a test of training lora with MOSS. I don't think it came out that great, so if anyone has training parameters that work well then I'm all ears.

Put "text, speech bubble" in negatives if you keep getting those. Works well on 0.8 and 0.9 weights. Mostly good for closeups. It seems to like "hand drawn colored sketch style" to mostly match.

"Dynotaku" for trigger.