Xinhua , Chinas state-run press agency, has unveiled new AI anchors — digital composites created from footage of human hosts that read the news using synthesized voices. Its not clear exactly what technology has been used to create the anchors, but theyre in line with the most recent machine learning research. It seems that Xinhua has used footage of human anchors as a base layer, and then animated parts of the mouth and face to turn the speaker into a virtual puppet. By combining this with a synthesized voice, Xinhua can program the digital anchors to read the news, far quicker than using traditional CGI. (Weve reached out to AI experts in the field to see what their analysis is.) According to reports from Xinhua and the South China Morning Post, two anchors (one for English broadcasts and one for Chinese) were created in collaboration with local search engine company Sogou. Xinhua says the anchors have endless prospects and can be used to cheaply generate news reports for the agencys TV, web, and mobile output. Each anchor can work 24 hours a day on its official website and various social media platforms, reducing news production costs and improving efficiency, says Xinhua. The technology has its limitations. In the videos above and below of the English-speaking anchor, its obvious that the range of facial expressions are limited, and the voice is clearly artificial. But machine learning research in this area is making swift improvements, and its not hard to imagine a future where AI anchors are indistinguishable from the real thing. This will strike many as a disturbing prospect, especially as the technology is being deployed in China. There, the press is constantly censored, and it is nearly impossible to get clear reports of even widespread events like the countrys suppression of the Muslim Uighur community. Creating fake anchors to read propaganda sounds chilling. But what the actual effect on society may be if such anchors become widespread is hard to judge. If Xinhua wants someone to read the news without questioning it they dont need AI to make that happen. Meanwhile, synthetic characters are slowly finding their way into mainstream culture, with figures like virtual pop star Hatsune Miku and CGI Instagram models familiarizing the public with this sort of creation. But while these examples fall clearly into the world of entertainment, having AI anchors read the news suggests the technology could become more than a novelty.
Theyre modeled after human anchors and are animated to speak. China's state-run news agency Xinhua has unveiled the latest additions to its team of reporters -- two AI anchors. The two anchors, one that speaks in English and another in Chinese, have the likeness of some of Xinhua's human anchors, but their voices, facial expressions and mouth movements are synthesized and animated using deep learning techniques. " AI anchors have officially become members of the Xinhua News Agency reporting team," the agency said. "They will work with other anchors to bring you authoritative, timely and accurate news information in both Chinese and English." China's South China Morning Post reports that the AI anchors are available through Xinhua's English and Chinese apps, its TV webpage and its WeChat public account. The technology behind the anchors is being provided by search engine company Sogou. Xinhua says its AI anchors can deliver the news with the "same effect" as that of human reporters. But if you watch the video, that isn't exactly true. It's pretty clear you're watching a non-human anchor as the mouth movements and facial expressions aren't quite human-like, and the voice can come off as a little robotic. But as recent research has shown, manipulating video to make images say what you want while still looking shockingly natural, is within reach. And maybe future AI anchors, if there are more, won't be stuck in the uncanny valley. Unlike human anchors, Xinhua's AI anchors can work around the clock, and the news agency says they, therefore, allow for improved production efficiency and reduced costs.