@ianfontinell I don’t know if you’ve watched, but let’s use codynextdoor for example. A few continuity errors off the top of my head are:
face / body shape changing: Rick / Cody’s dad can be really buff one scene and lot less in the next one when he wears clothes. The mom looks older with short hair and when she wears the nurse outfit she’s a lot younger with longer ponytail, looking like a completely different person (but who cares about her, right? Though, poor her for having Rick as a husband. The guy is a cheating creep)
environment: the shower scene after dodge ball game, notice how the shower stall is different each shot. Then the private sauna room, at one point the sex moved from the seat to literally on top of the hot coals Also, a bedroom scene from episode 2 suddenly adds an extra bed in front of the one Rick and Cody is on (nonsensical background is pretty common in Stable Diffusion too, because these AIs can’t actually think)
Probably just being too nitpicky with those and they don’t bother me as much as appearing/disappearing clothes, but they’re fun to notice.
I won’t say I’m that knowledged, but I assume for video generation models they must've fed actual porn footages into the training data considering how good the close up penetration scenes are, but not enough to get the erections consistent. To be fair, a lot of image generation models also struggle with getting erect cock to look right without the help of LORAs, and those are always made by the open sourced community like Civitai, not the company who made the models.
