Ey, thanks for your comment.
Well.. I'm biased (obviously) but I'm pretty convinced that traditional work will never go away. First, the obvious, how do you think AI is going to "learn"? There are tons of models available which you can use to set up sceneries, but all that data comes from.... you guessed it: images and media which were all made "the old fashioned way". Well, maybe not all but you'll quickly notice when you're fed with artificial AI "dribble data" vs. something more realistic.
And then there's the NSFW part. We all know that sex sells, and that also holds true for this free AI platform. If you check out the so called "model data" you'll come across dozens of (humanoid) models which clearly mention their usability for NSFW as well. In other words: ask for a naked girl or guy and you'll be looking at curves or muscles in no time.
Yah, but there's anything that Daz3D taught us it's that designing a humanoid figure is one thing... but the anatomy? That's a whole other ballgame. And even then there's anatomy... and then there's actually usable anatomy. I enjoy the official penis & vagina props for G8F, but they're no match for things like dicktator, cock-o-rama, the advanced pussy and of course the golden palace.
Guess what? You see the exact same things happening in the world of AI. Even if you have a model that was obviously meant for NSFW then there's only so much between the legs. Want more? You'll need extra filters, which of course aren't necessarily fully compatible with your model. Public hair? What's that?
And then we get to the naughty fantasies... sex, having your model tied up? Ropes kinda work, chains become weird.
Don't get me wrong here: there are models ("merge checkpoints") which are specifically geared towards pr0n. A woman sucking a dick for example. But how about some casual sex?
Or imagining a snake coiling around a figure? Or how about a humanoid figure stuck in spider webbing? Lo and behold: while there are some extra 'filters' ("LoRA's") to make that work, most of those are geared towards cartoons. How about a monster grabbing someone... that will immediately become a challenge.
And then you're suddenly realizing that you're working with a data collection that's an easy 120Gb+ large (still images only), yet can't do so much as simulating simple sex, or wrap a snake around a (naked?) humanoid figure. And how long has AI already been available?