Chinese companies develop AI video generation technology

BEIJING, Aug. 3 (Xinhua) — Several major artificial intelligence (AI) video generation models developed by Chinese technology companies have been made available to users around the world, acting as Chinese equivalents of OpenAI’s Sora.

The latest version of the tool, called Vidu, developed by Chinese artificial intelligence company ShengShu Technology and Tsinghua University, became available globally on Tuesday and offers text-to-video and image-to-video generation capabilities.

The advanced AI model, unveiled in April at the Zhongguancun 2024 Forum in Beijing, now allows users to create a four-second or eight-second video in 1080p high definition with a single click, said Tang Jiayu, co-founder and CEO of the company.

Tang said such technology would revolutionize the content industry, including the film and television animation sector.

“Over the next decade, the barriers and costs of creating and producing films and television programs will be further reduced, and imagination and creativity will be better transformed into productive forces,” he said.

This isn’t the first Chinese AI video generator to launch globally. Last Friday, Beijing-based tech firm Zhipu AI launched its large-scale video generation model, known as Ying.

As Zhang Peng, CEO of the tech company, said, the product can generate videos not only from texts and images but also from video clips.

In June, short video platform Kuaishou also launched a large video generation model called “KLING”, which enables the creation of realistic motion scenes and accurate physics simulations when generating text on video and image on video.

There are already several technology companies around the world working on text-to-video and image-to-video generation models. In China, the main AI video generation products adopt a framework that combines a deep learning model with a diffusion model, relying heavily on algorithms, computing power and data, according to Tang and Zhang.

Zhang admitted that the domestic AI video generation technology is still in its infancy and imperfect, and a number of problems, such as high-quality data and algorithms, need to be solved.

“The best way to accelerate the development of artificial intelligence is to conduct parallel research, development and application of the technology, rather than keeping the technology locked away in a lab and developing it in isolation,” he said.

So far, China has submitted and launched more than 180 AI-based generative content models that can provide public services, and the number of registered users has exceeded 564 million, Wang Jingtao, deputy director of the Cyberspace Administration of China, said at the 12th Internet Security and Artificial Intelligence Conference in Beijing on Wednesday.