How do you all think about this? Are you actually using SDXL text2img workflows for client-ready cases, or do you stick to MJ and similar tools when you need ultra sharp, realism, sharp, on-brand visuals?

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. Then make em kiss