Tag
OpenAI's technical report on Sora describes a video generation model that unifies diverse visual data through visual patches, enabling large-scale training of generative models capable of producing high-definition videos up to one minute long across variable durations, aspect ratios, and resolutions.