Deepfakes Of Chinese Influencers Are Livestreaming 24/7
Once the avatar is created, the mouth and body move to the rhythm of programmed sound sequences. While scripts were once written in advance by people, companies now use large language models to create them.
Now, all human employees need to do is enter basic information like the brand name and price of the product being sold, edit the generated script, and watch the digital influencer take action. A more advanced version of the technology can display comments in real time and search its database for relevant responses in real time, making it appear as if the AI streamer is actively interacting with viewers. According to Seema, the company can also adapt its marketing strategy based on the number of viewers.
These streaming AI clones will be trained on common scripts and gestures seen in e-commerce videos, said Huang Wei, head of live streaming at Chinese AI company Xiaoice. The company has a database of almost hundreds of ready-made events.
“For example, [streamers] say, 'Welcome to my live channel.' You move your fingers and click the follow button.” They're definitely pointing out why the 'subscribe' button is on the screen of most mobile streaming apps ,” says Huang. Similarly, when streamers promote a new product, they point to a shopping cart where viewers can access all the products. Xiaoice AI Streams replicates all of these methods “We want to make sure that spoken language and language of the body match. You don't I want you to clap and talk about the "subscribe" button. It seems strange," he says.
Chen says he's noticed a lot more interest in AI feeds from brands this year, in part because everyone wants "降本增效," a new buzzword among Chinese tech-related companies as the national economy slows down.
Chen now has more than 100 customers using Xiaoice's services, and sales from these virtual streams have generated millions of dollars. A performance by Xiaoice raised more than 10,000 yuan ($1,370) in one hour.
He said there were still weaknesses. Many of its customers are furniture brands, for example, and while the AI is smart enough to talk and use gestures, it can't actually sit on a couch or lie on a bed, making for engaging discussions to test by real users. . Some products.
In addition to small startups like Silicon Intelligence and Xiaoice, large tech companies are also experimenting with AI-powered live streaming. Alibaba, Tencent, Baidu and JD have launched several similar services this year that allow brands to generate their own AI feeds on their platforms.
