Human faces were excluded on purpose from the DALL-E 2 training set in order to prevent misuse. I suppose the same will be the case here (or at least, in public

Human faces were excluded on purpose from the DALL-E 2 training set in order to prevent misuse. I suppose the same will be the case here (or at least, in public-facing versions of the models). Given how it renders dog faces, I don't see why it wouldn't be good at human faces too, if trained for it.