Pictured present is an AI-generated clip from Vidu's website. The instrumentality tin make videos from substance oregon representation prompts.
Evelyn Cheng | CNBC
BEIJING — Beijing-based Shengshu Technology connected Wednesday said that its artificial intelligence-powered text-to-video instrumentality Vidu volition present beryllium capable to make videos by combining images.
Vidu already allows users worldwide to make 8-second clips based connected written prompts. While OpenAI — the shaper of ChatGPT — successful February revealed that its AI exemplary Sora could make one-minute videos from text, it has yet to merchandise that publicly.
Vidu's caller AI diagnostic tin harvester 3 pictures — specified arsenic a shirt, idiosyncratic and moped — into a video of the idiosyncratic wearing the garment and driving the moped done a scene, Shengshu said.
Other platforms assertion they tin crook substance oregon images into videos utilizing AI, but the prime of output varies. The breakthrough that Shengshu claims is the quality to instrumentality 3 unsocial images and integrate them with ocular consistency into an AI-generated video.
"Very aboriginal connected we pinpointed [visual consistency] arsenic the problem, and wanted to lick it well," Fan Bao, main exertion serviceman astatine Shengshu, said successful Mandarin, translated by CNBC.
Vidu launched successful April and its quality to crook 2 illustration photos into lifelike videos of radical hugging went viral connected TikTok.
The AI video generator is already making wealth from advertisers, animators and different businesses, Shengshu co-founder and CEO Jiayu Tang said successful Mandarin, according to a CNBC translation. He said monthly usage rates per lawsuit tin scope from 100,000 yuan to 1 cardinal yuan ($13,871 to $138,711).
To code copyright issues, Tang said a institution mightiness motion a woody with an creator that allows the AI to mimic the artist's benignant of coating for an advertisement. He said helium hadn't seen important ineligible cases astir consumers' usage of images.
Tang added that Vidu doesn't let the nationalist to make contented utilizing images of celebrities oregon "sensitive" individuals. He said the AI instrumentality besides bans nudes and convulsive images. As for idiosyncratic photos, Tang said Vidu destroys the information successful accordance with wide information extortion regularisation — a planetary benchmark.
Shengshu was founded past twelvemonth with backers including Baidu Ventures, Alibaba-affiliate Ant Group, Chinese startup Zhipu AI, Qiming Venture Partners and Beijing city, according to PitchBook.
Tang said Vidu's AI runs disconnected rented unreality servers successful China and abroad.
English (US)