One email each morning — yesterday's AI, sortedGet it in your inbox

METAL LAB

Synthesia releases 'hard mode' quiz video testing real vs. AI footage

The AI avatar video company put viewers' ability to spot generated video to the test on YouTube Shorts

이미지: METAL LAB 생성

Summary

  • Synthesia posted a "hard mode" quiz on YouTube Shorts asking viewers to distinguish real footage from AI-generated video
  • No additional details, such as specific figures or correct-answer rates, were disclosed beyond the video itself
  • The move is seen as illustrating how AI-generated video has reached a level of quality that is difficult to identify by eye
Video from the source
게시자
신테시아(Synthesia) 유튜브 채널
형식
유튜브 쇼츠, 실사·AI 영상 구별 퀴즈
제목
Can You Guess Real vs AI Videos? (Hard Mode)
발행일
2026년 8월 10일

Real or AI? Even the eye can't tell

Synthesia posted content on YouTube Shorts in the form of a quiz, lining up several short clips and challenging viewers to identify which were real footage and which were AI-generated. The word "hard mode" in the title suggests the difficulty has been raised compared to previous versions. However, the released material contained no additional details beyond the video itself, such as the correct-answer rate or the names of the generative models used.

What kind of company is Synthesia

Synthesia is an AI avatar video generation platform that lets users create videos featuring people using only text input. Without filming an actual actor, it can produce footage that looks like a specific person speaking, making it widely used for corporate training materials, internal announcements, and marketing videos. It's notable in itself that a company like this would put out content daring viewers to "tell it apart from real footage." The move both promotes the sophistication of its own technology and reflects a broader industry trend in which generated video has become increasingly hard to distinguish from real footage.

AI video generation technology has advanced in recent years to the point of precisely reproducing facial expressions, lip movement, lighting, and even camera shake. In the early days, AI-generated video could be easily spotted through flaws like unnatural blinking or distorted fingers, but such defects are rapidly disappearing. As a result, it is becoming an increasingly difficult task for ordinary viewers to distinguish real footage from generated content just by watching a short clip.

So what does this change

The rise of this kind of content speaks to two things at once. First, AI video generation companies are actively using "spot the difference" quizzes as a marketing tool to visually demonstrate their technical capabilities. Second, it signals that the very standards viewers use to judge the authenticity of online video are being shaken. For content where trust matters most—news footage, interviews with public figures, product demonstrations—this is a moment that calls for a stronger habit of verifying the source and method of production.