But I have not tried making generative models with out-of-distribution data before. Distributions other than main training data.
There are several indie attempts that I am aware of. Mentioning them to the reply of this comment. (In case the comment gets deleted)
The first layers should be general. But the later layers should not behave well to porn images. As they are more specialist layers learning distribution specific visual patters.
Transfer learning is posssible.
2. hentAI automates the process: https://github.com/natethegreate/hent-AI
3. [NSFW] Should look at this person on Twitter: https://twitter.com/nate_of_hent_ai
4. [NSFW] PornHub released vintage porn videos upscaled to 4k with AI a while back. The called it the "Remastured Project": https://www.pornhub.com/art/remastured
5. [NSFW] This project shows the limit of AI-wthout-big-tech-or-corporate-support projects. This project creates female genitalia that don't exist in the real world. Project is "This Vagina Does Not Exist": https://thisvaginadoesnotexist.com/about.html