Deepfake

Deepfakes are made by training a model on footage of a person so it can generate new frames or speech in their likeness. Quality has moved from obvious to genuinely hard to spot in a few years, and the tools need less source material than they used to, which matters for anyone with hundreds of hours of their own face online. The common version creators meet is a scam advert using a familiar face to sell crypto or a fake giveaway. YouTube's responses are likeness detection, which lets creators find unauthorised synthetic uses of their face and request removal, and required disclosure when a video contains realistic synthetic content. Neither stops the video existing; both give you a route to act.