I wonder how long until we can also take drawings/characters and combine them with porn videos to get a duplication of the whole pose/actions with the animated characters doing them.
You can probably do that now with AnimateDiff. I haven't tried because my graphic card isn't powerful enough.
A text2video model like Kling, but trained on porn and summertime saga, would output porn without needing a base video. We need people with tons of money that aren't moralist fuckers to train these models, though. Unless we can train adapters like with do with stable diffusion.
Imo, what we'll be seeing next are multimodal models that will output a 3d representation in addition to the 2d video. You'll be able to load the scene in a 3d program, animate it, edit objects. The movable objects will even be capable of being rigged. It's going to be crazy.
