Video generation models as world simulators
We explore large-scale training of generative models on video data. Specifically, we train text-conditional diffusion models jointly on videos and images of variable durations, resolutions and aspect ratios. We leverage a transformer architecture that operates on spacetime patches of video and image latent codes. Our largest model, Sora, is capable of generating a minute of high fidelity video. Our results suggest that scaling video generation models is a promising path towards building general purpose simulators of the physical world.المصدر: OpenAI Blog | Source: OpenAI Blog
ملاحظة تحريرية | Editorial Note: نُشر هذا المقال في الأصل بواسطة OpenAI Blog. خبر (Khabr) هي منصة إعلامية أردنية مرخّصة تعمل بالذكاء الاصطناعي. نضيف قيمة تحريرية من خلال: تحليل ذكي للأخبار، ملخصات تلقائية، رواية صوتية بالذكاء الاصطناعي، ترجمة متعددة اللغات، وتدقيق الحقائق. هدفنا جعل الأخبار أكثر وضوحاً وسهولةً للقارئ العربي.
This article was originally published by OpenAI Blog. Khabr is a licensed Jordanian AI-powered news platform (Registration #82086). We add editorial value through: AI-powered news analysis, automated summaries, AI audio narration, multi-language translation (Arabic, English, French, Turkish), and AI fact-checking. Our mission is to make news more accessible and understandable for Arabic-speaking audiences worldwide.
