Sora’s Got Competition: China’s Peking Uni Hustling To Mimic OpenAI’s Masterpiece

In a bid to replicate OpenAI’s revolutionary text-to-video generator, Peking University has joined forces with RabbitPre, an AI company based in Shenzhen, China.

What Happened: The project’s GitHub page reveals that a small team of researchers from Peking University is striving to recreate Sora, OpenAI’s trailblazing text-to-video generator, as reported by TechNode on Tuesday. The original model, launched in February, has pushed the boundaries of GenAI, leading to a surge in interest and activity in the field.

Due to a lack of video data, the team is reaching out to the open-source community for resources, despite being unable to complete full training. They have, however, devised a three-part framework, including a Video VQ-VAE, a Denoising Diffusion Transformer, and a Condition Encoder. Their initial progress, featuring four reconstructed video demos of various resolutions and aspect ratios, has been shared on GitHub.

Why It Matters: The launch of OpenAI’s Sora in February marked a significant advancement in AI technology. The model’s ability to generate high-quality videos from text prompts has been widely lauded. However, as noted by tech influencer Marques Brownlee, the model still hasits limitations, suggesting that there is room for improvement in the technology.

For more context, OpenAI CEO Sam Altman has expressed a strong focus on developing artificial general intelligence (AGI), following the successful debut of the Sora AI model. This indicates the broader implications of advancements in AI technology, such as the development of Sora, and the potential impact on the future of AI.