Ankündigung des Herstellerserreichbar
Video generation models as world simulators
Der technische Bericht vom 15.02.2024, in dem Open
geprüft 10.11.2026
Worauf sich diese Seite beruft, wörtlich, abgerufen am 05.09.2026:
Our results suggest that scaling video generation models is a promising path towards building general purpose simulators of the physical world.
bestätigt 10.11.2026Whereas LLMs have text tokens, Sora has visual patches.
bestätigt 10.11.2026We find that video models exhibit a number of interesting emergent capabilities when trained at scale. These capabilities enable Sora to simulate some aspects of people, animals and environments from the physical world.
bestätigt 10.11.2026Sora currently exhibits numerous limitations as a simulator. For example, it does not accurately model the physics of many basic interactions, like glass shattering.
bestätigt 10.11.2026Model and implementation details are not included in this report.
bestätigt 10.11.2026Importantly, Sora is a diffusion transformer.
bestätigt 10.11.2026At inference time, we can control the size of generated videos by arranging randomly-initialized patches in an appropriately-sized grid.
bestätigt 10.11.2026Our largest model, Sora, is capable of generating a minute of high fidelity video.
bestätigt 10.11.2026Likewise, it can generate multiple shots of the same character in a single sample, maintaining their appearance throughout the video.
bestätigt 10.11.2026
Text-zu-Video im GlossarZeitliche Beständigkeit im Glossar09 Weltmodelle13 Video, Musik und 3D