Nearly all large AI models currently come with more or less usage restrictions, and adult-oriented content in particular is strictly prohibited.
I might build a small AI server of my own once the cost of relevant hardware drops in the future, which I could use for animation rendering. Obviously, this is not feasible right now due to the exorbitant hardware prices.
Actually if you call through API instead of browser chat bots and desktop apps, the censorship is much more muted. I know at least Deepseek has no problems handling NSFW written material, so that should help greatly in AVN coding. As for animations, yea still no way to generate adult stuff on mainstream models to date.
You might be better off just setting up your workflow on something like ComfyUI cloud and just pay for server usage rather than buying exorbitant GPUs like 5090 which is so ridiculously overpriced that you'l need to generate tons of animations to break even.
Even the best 5090D can only host at most a heavily quantisized ~30B model, so it's gonna suck badly at agent work, not to mention much slower. It makes more sense to subscribe or buy tokens on the go IMO. Just give it a try, but $5 worth of tokens from Deepseek or Mimo (2 cheapest mainstream models) and give it a try.
I believe personal AI hardware devices will become far more widespread and affordable in the future.
There are already personal AI servers available that I can afford right now.
Using ComfyUI, I can generate a 10-second 1080P animation within ten minutes,
though the results still haven’t lived up to my ideal expectations.
Since over two-thirds of the entire production workflow has already been completed,
no further revisions will be made in theory. All upcoming projects will still focus primarily on static imagery.
Animation features may not be incorporated until my next work,
by which time AI animation production should be much more refined and polished.
