Adobe VoCo allows users to capture audio as a text file and then alter the text to change the words that the person has spoken. It can then play the altered text back in the original voice, making it sound like the person is actually saying whatever is in the text file. The app needs about 20 minutes of training (which could be done via a speech on youtube, for example) to learn a person's "voice".
Face2Face allows for real-time facial manipulation of an existing video. The user raises an eyebrow, the face on the video raises an eyebrow. The user says something, the face on the video moves their lips as if they are saying it too.
These apps are still in their infancy, but once they are perfected, it's going to get very difficult to determine what's real and what has been generated by the combined apps.