From South China Morning Post:

Chinese researchers from Peking University and Rabbitpre launch the Open-Sora plan to develop a text-to-video model similar to OpenAI’s Sora. They aim to create a simple version with help from the open-source community. Tencent AI and ByteDance also released similar models limited to two-second videos.

The team behind Open-Sora has developed a framework and demoed video reconstructions of various lengths and resolutions. They plan to enhance resolution, train with more data and GPUs. OpenAI’s Sora demo videos have sparked interest in China, but some express concerns amid US trade restrictions.

Rabbitpre and Peking University collaborate on the Open-Sora project, with 13 initial team members including professors and company executives. The joint lab is dedicated to AI content research, highlighting China’s push in generative AI technology. Other companies like Tencent AI and ByteDance have also released text-to-video models with limitations on video length.



Read more at South China Morning Post: Chinese researchers aim to ‘reproduce’ OpenAI’s text-to-video model Sora with help from open source community