>>5413267I just started using it today. You need to figure out a "prompt" that works for you, and usually set it to give back 6 images at a time. 90%+ are duds, but you just refine your prompt to get a better image.
What I recently found is that using a pre-existing image to start with can help the AI greatly in some ways. For some reason, it is crazy good with faces, but it is the worst with humanoid bodies.
I cannot get space elevators proper into the artwork yet, so I will keep trying that.
Some people on /g/ have crazy refined prompts and fine tuning.
