I'm working on a project where I want to generate images featuring three consistent characters: two men and one robot. I’ve trained custom LoRAs for each of them using Flux. Right now, my workflow looks like this: I generate an image using the robot's LoRA and just two random male characters.