Making an ‘AI human’… I started dreaming of producing news without an anchor.

I chose a person in his 40s with long black hair, a white face, and a low-pitched voice. Since you’re a reporter, why don’t you hold a piece of paper? We created ‘Artificial Intelligence Anchor’ with the idea of ​​creating a ‘virtual Hankyoreh reporter Lim Ji-seon’. After writing down the script for the anchor to read, preparations for creating the video are almost complete. After inserting the anchor background photo and slightly adjusting the anchor position on the screen, I pressed the ‘Create Video’ button. A 33-second news video was completed in just 3 minutes without hiring a newscaster. This month, KT unveiled the ‘AI Human Studio’, which allows users to create virtual humans and easily produce videos. Last year, Katie, who launched the ‘AI Voice Studio’ that can create an ‘artificial intelligence voice’ using 110 voice samples or one’s own voice, has gone one step further with a product that can easily create a ‘virtual human’ with a human shape and voice. A public platform has been opened. After accessing the studio website and registering as a member, you will see five templates depending on the type of virtual human you want to create, such as instructor, announcer, anchor, show host, and emcee. When you select ‘AI Anchor’, six male and female virtual human models of different genders and races appear. It is said to be a virtual model with no portrait or copyright issues, with a face created by artificial intelligence added to the appearance of an actual model. Once you’ve decided on your outfit and style with one click, all that’s left is to enter a news background photo and script. Pronunciation correction is also possible. If there is a word that the artificial intelligence anchor pronounces strangely, it can be corrected by writing it in the ‘Smart Vocabulary’ or it can be paused for about 0.2 seconds between sentences. ‘KT’ was pronounced as ‘KT’ and ‘service’ was pronounced as ‘service’ and it became much more natural. If you put your mind to it, you can embody five emotions (angry, sad, neutral, calm, happy) in the voice of an artificial intelligence anchor, and you can become a ‘talented person’ who speaks five languages ​​(Korean, English, Chinese, Japanese, and Spanish). It may be possible. Until now, the term ‘AI virtual human’ has been a method of producing specific people, such as ‘AI Kim Joo-ha’ and ‘AI Choi Gyeong-ju’, as artificial intelligence virtual humans for the needs of specific companies. However, the service released by KT through technical cooperation with CN AI, a company specializing in generative artificial intelligence, allows ordinary people to easily use virtual humans. The release of a paid version and the integration of Katy’s super artificial intelligence ‘Faith’ are planned in the future. We are also starting to produce ‘customized artificial intelligence humans’ for corporate customers. Customized production requires 4 to 8 hours of studio filming in which the costume is changed and the given movements are performed, as well as the process of accurately reading the 30 sentences provided. Katie said, “When creating an AI Human from a real human, we are making efforts to ensure safety by intentionally complicating the process.” Reporter Lim Ji-seon sun21@hani.co.kr