You can always tell it's a deep fake by looking at the neck. Where the CGI projection meets the body double whose face they're tracking.
I think it's easier for them to project a CGI deepfake onto a subject using facial/camera tracking instead of generating a whole entire body. If they use a double, they only have to focus on the face, head, and neck. A whole body would likely be much easier to detect as fake because it doesn't look organic.
not too far from me, kek