Chinese researchers from Peking University and AI company Rabbitpre have launched the Open-Sora project aiming to replicate OpenAI’s text-to-video model, Sora, in a more accessible and scalable way with community involvement. The project has made notable progress by developing a three-part framework and showcasing four video demonstrations of varying resolutions and durations. Future goals include enhancing technology for higher-res videos and bolstering training with more data and GPUs to boost computing power.
The emergence of OpenAI’s Sora has sparked interest and concerns within China’s business and tech sectors. While some companies are eager to leverage text-to-video AI, others worry about China’s competitiveness due to US export restrictions on high-performance GPUs. Major Chinese tech firms like Tencent AI and ByteDance have also entered the text-to-video arena with projects like VideoCrafter2 and MagicVideo-V2.
The collaborative Open-Sora initiative between Peking University and Rabbitpre has established the Rabbitpre AIGC Joint Lab dedicated to AI content research. The team, led by experts like Assistant Professor Yuan Li, Professor Tian Yonghong, Rabbitpre’s Dong Shaoling, and CTO Zhou Xing, aims to push the boundaries of AI-generated video content. This project represents China’s latest move in the ongoing AI race, with its success poised to influence the future landscape of artificial intelligence.
HEALTH
WORLD AFFAIRS
TECHNOLOGY
CREATIVE
