I understand why generated images and video of Scarlett Johansen always looks the same since AI has been trained on millions of videos of her. But how would they generate the same AI "actress" in each generated video/image when there's nothing to train on since that actress wouldn't appear in the training data? Also, what will be the difference between an AI actor/actress acting as a character and an AI just being told to be a character (like in AI roleplay)?
[link] [comments]