I guess you're simply using Stable Diffusion v1.4/1.5, you can improve the quality of the model if you finetuned it on high quality pornographic images, if NovelAI images are sometimes hard to distinguish from actual real images, than you can also do it if you finetuned it.
I've finetuned the model on pictures, but I'm lacking proper dataset. I've annotated some images using scripts - but those are just tags
I should maybe make it clearer of a prompt schema -it works much better on tags and then small description
You should add a share button. Some generations are hilarious