Raw LLM Responses

Yeah. The Obama thing also required more effort too. We're somewhere between months to years away of someone generating something many times more convincing than that video at a fraction of the effort. Hypothetically I'll be able to soon record a speech with a voice changer to make it basically perfect Obama with emotional range, then feed that audio into an AI video generator with a basic description to describe what the video should be and then it'll generate the footage from scratch, perfectly synchronized to the provided audio. The above (providing audio and generating a synchronized video) is a solvable problem. Detecting its fake with absolute certainly is not.