The project was not launched at the state level
A couple of weeks ago, OpenAI once again showed how impressive the pace of AI development can be by introducing the Sora neural network, which can generate videos from a text description. It is reported that a project has already been launched in China to »reproduce» this AI.
Professors from China's Peking University and artificial intelligence company Rabbitpre have jointly launched the Open-Sora project, which aims to «replicate OpenAI»'s video generation model. The authors want to create a simple and scalable version of Sora, and based on open source code. The project has also been launched on GitHub.
The team had already developed a three-part framework and showed four demos of reconstructed videos in different resolutions and aspect ratios, ranging in length from 3 to 24 seconds. The team's future goals include fine-tuning to achieve higher resolutions, as well as training using more data and more GPUs.