Would using the words “pre-hormone, pre-surgery, feminine-presenting trans man” in the prompt count of the output is undistinguishable from a woman?

The more that I understand how model training works. The more that I feel that it's almost certain. Since, I don't believe the major labs are likely to filter everything NSFW out of the pre-training data and instead will rely on post-training to accomplish avoiding NSFW outputs.

They will probably not feed it any pornographic data, what you mean by NSFW?

Mild eye roll to the threshold but that's roughly "Industry standard" so It's understandable. It's been easy enough to adapt prompts cleverly for every other image generator, I don't see why the latest and greatest will be any different.

I have another (more popular) market about more explicit pornography. I'm not intending to imply any moral judement either way.

@jacksonpolack lol I'm adding commentary for the sake of discussion. I have to laugh because by this time next year I'll probably go from instagram SFW to instagram NSFW if I post shirtless (see my market, I can delete embed if it's unwelcome)