I think you are really underselling ControlNet. I was able to get it making exactly what I wanted extremely fast through the Automatic1111 interface. And that's after only making AI art for a week before installing it. The best part is that it doesn't require any proprietary software like Photoshop.
ControlNet solved the composition problem, the pose problem, and if you have a 3d modeling program to pose the hand, the hand problem. The key is to use 2-3 controlnets at once for the initial image, and then inpaint where there are problems.
I think you are really underselling ControlNet. I was able to get it making exactly what I wanted extremely fast through the Automatic1111 interface. And that's after only making AI art for a week before installing it. The best part is that it doesn't require any proprietary software like Photoshop.