OpenAI unveiled the text-to-video model Sora, capable of generating detailed videos from simple text prompts. The AI firm acknowledged that while Sora can create intricate scenes with multiple characters and emotions, it still struggles with accurately simulating complex physics in scenes. Built on past research models like ChatGPT and Dall-E 3, Sora operates on a diffusion model to gradually transform static noise into coherent videos up to 1080p resolution. Despite its strengths, OpenAI admits that Sora may mix up spatial details or generate physically implausible motions due to weaknesses in simulating cause and effect accurately.
翻譯成其他語言
從原文內容
cointelegraph.com
從以下內容提煉的關鍵洞見
by Tom Mitchelh... 於 cointelegraph.com 02-26-2024
https://cointelegraph.com/news/sora-openai-video-generation-model-artifical-intelligence-weakness深入探究