Known formally as Beijing Zhipu Huazhang Technology, the unicorn started to develop its own text-to-video large language model (LLM) before OpenAI revealed video clips made by Sora in February, according to a report from TMTPost on Monday.
Zhipu AI is expecting an explosion of LLMs capable of turning text prompts into videos this year because of growing demands from a diverse array of customers with needs ranging from film to video game production, the report said, citing the company.
The Beijing-based start-up aims to release “high-quality text-to-video tools” this year at the earliest, according to the report.
Zhipu AI did not immediately respond to a request for comment on Tuesday.
Chinese companies have been racing to catch up with world-leading generative AI players ever since Microsoft-backed US start-up OpenAI introduced the ChatGPT conversational bot in late 2022, followed by Sora in February this year.
The AI model can produce videos with 1080p resolution that are no longer than 16 seconds, based on simple text prompts, according to Shengshu. By comparison, OpenAI said Sora can generate videos up to 60 seconds long.
However, OpenAI’s services are not officially available in China, where around 200 LLMs – the technology underpinning generative AI services – have sprung up, according to recent government figures.
Zhipu AI counts venture capital firms and domestic tech giants as backers, including Tencent Holdings, Meituan, Xiaomi and Alibaba Group Holding, which owns the Post. Last year, it raised 2.5 billion yuan (US$346 million) from investors.