Home>Updates

AI virtual anchor debuts

chinadaily.com.cn | Updated: Nov 8, 2018 L M S

The world's first human replica intelligent virtual host, AI virtual anchor debuts on Wednesday at the fifth World Internet Conference in Wuzhen, Zhejiang province.[Photo provided to chinadaily.com.cn]

On Nov 7, the fifth World Internet Conference kicked off in Wuzhen, Zhejiang province. At the conference, Wang Xiaochuan, CEO of Sogou, Liu Siyang, vice president of Xinhua News Agency, and famous TV host Qiu Hao launched a distinctive product: the AI virtual anchor---the world's first human replica intelligent virtual host.

Once a user inputs a news text, a virtual Xinhua news anchor will appear on the screen. The host can not only broadcast with the same voice as the real person, but his lip movements and facial expressions can be matched perfectly with his voice. The virtual anchor is not much different from a real one.

According to sources when developing the virtual anchor Sogou technicians worked closely with the Xinhua news anchor.

Based on "Sogou avatar" technology and using such cutting-edge technologies as facial landmark localization and face reconstruction, the AI Virtual Anchor was developed successfully side-by-side with multi-modal information for joint modeling training.

According to Wang Yufeng, general manager of the intelligent voice division at Sogou, “Sogou avatar” technology is one of the core achievements of Sogou artificial intelligence, which follows the concept of “Nature Interaction + Knowledge Computing”.

This form of broadcasting breaks through the restriction that virtual images must be created first then be matched with voice. Now, using the "Sogou avatar" technology, the AI Virtual Anchor can produce synchronized video in real time.

After users provide text through various ways such as text typing, voice input and machine translation, they instantly obtain a real-time broadcast video. This way of newsmaking will greatly reduce the costs of post production, and improve its efficiency.

“Sogou avatar” couldn’t demonstrate such a remarkable performance without years of technical exploration and continuous research and development. As early as 2012, Sogou began researching speech technology, and gradually developed China's largest voice recognition engine.

Currently, there are 500 million voice requests of Sougo each day. It handles them with multi-lingual and multi-tone speech synthesis capabilities, and can realize personalized voice synthesis and emotional transference.

The technical team that developed the AI virtual anchor has won many championships at international top machine vision events like IEEE CVPR WAD, Pascal VOC and Mega Face as well as Blizzard Challenge competitions, the most prestigious award in the field of speech synthesis.

"Sogou avatar" technology realizes more natural human-computer interaction in different scenarios. In addition to its application in the field of media convergence, it can offer personalized production in various fields such as entertainment, medical health, education and law.

It can be anticipated that this new technology will improve social productivity and service efficiency, reduce production costs of industries, and enhance people's experiences in science and technology.