[The performance of the new AI model in video conversations is almost realistic] On October 9th, the official website of Tavus, an AI video company in San Francisco, USA, stated that the company’s latest new model Griffin-Lite organized 54 subjects into a one-minute video call, and 26 people mistakenly thought that the other party was a real person, accounting for 48%. Under the same test process, only 2.4% of the subjects judged this way for the previous generation system. Tavus claims that this is the first real-time video conversation model to pass the "Video Turing Test" and defines it as a "human interaction model." Unlike traditional voice assistants, Griffin-Lite can respond in real time when both parties are speaking at the same time, whether it is echoing or interrupting, just like a real person chatting.