A CNN article highlights how to distinguish deep fake videos from real
ones. They suggest checking for specific natural details, spotting
potential mismatches, and watching out for clues in the behaviour of the
person filmed or background. This can offer interesting insights for VDA researchers who use “ready-made” video data from online sources (i.e., video data they did not capture themselves, but which instead were uploaded online).

You can find the article here:

https://edition.cnn.com/interactive/2020/10/us/manipulated-media-tech-fake-news-trnd