>>15378
> […] the model […] is usually trained on a safe image base, […] even on your own pc, […] you cannot really do many things.
Yes, this is an additional obstacle. And there might be some more limitations by your hardware. I assume you don't have your own supercomputer with endless memory? Three minutes for a picture like the one you posted are an amount of time one can live with. But even with the file space and the memory that is available today in the consumer section your PC will be way behind that what a company can afford. Won't it?

> […] train my own [model] but it's going to be very time consuming, especially since I don't have a good dataset of correctly tagged guro images.
Is there a way to merge some models? If a direct solution is not possible, may be it can be achieved by taking a detour via generating "pseudo topical" training images. I am thinking of feeding the same prompts to different models to get similar poses, e. g. a nude woman sitting on a chair, a (probably dressed) loli sitting on a chair and a headless corpse sitting on a chair. These would be used to train a new model.