OpenAI Releases Sora Video Generation Model, Proposing That Video Models Can Serve as World Simulators
Video generation models as world simulators
AI BriefingOpenAI released the video generation model Sora, which can generate high-fidelity videos up to one minute long. Based on the Transformer architecture, the model processes spatiotemporal patches of latent video and image encodings, and jointly trains a text-conditioned diffusion model on videos and images with variable durations, resolutions, and aspect ratios. OpenAI believes that scaling video generation models is a promising path toward building general-purpose simulators of the physical world.
Source: OpenAI: Official Website Updates · openai.com