We have models now that will generate not-quite-convincing video of a talking head in real time on a consumer GPU. Given the pace of developments in the field,

We have models now that will generate not-quite-convincing video of a talking head in real time on a consumer GPU. Given the pace of developments in the field, I think it's utterly ludicrous to suggest that getting this type of model from "not quite right" to "good enough to fool a lot of people" will take decades, when so many other types of model have achieved that feat in a matter of years or months. https://humanaigc.github.io/chat-anyone/ https://humanaigc.github.io/chat-anyone/