THE MAGIC OF TEXT-TO-VIDEO Another great leap for creative Al  

AI魔法——文本生成视频 创造性人工智能的又一次飞跃

作　　者：David Morton 

机构地区：[1]不详

出　　处：《空中英语教室（中级版）》2024年第9期13-15,49,56,共5页Studio Classroom

摘　　要：In March of 2023,a notoriously bizarre video of Will Smith eating spaghetti was created using an early text-to-video Al tool called ModelScope and released on Reddit,a popular content-sharing platform.The technology,while impressive,was deeply limited in its ability to accurately represent the text prompt-some users even described the video as"nightmarish."Little did they know that a year later text-to-video Al would be creating videos so realistic that they would blurthe line between the digital and the real.2023年3月,出现一段怪异的视频,内容是威尔·史密斯吃意大利面的场景。视频由名为“魔搭”的早期文本生成视频AI工具创造而成,然后发布在热门内容分享平台Reddit上。文本生成视频技术虽然让人印象深刻,但在精确呈现文本提示方面的能力极为有限——有些使用者甚至把该视频描述为“梦一般”。殊不知,才过了一年,文本生成视频的AI技术已经能创造出极为逼真的视频,足以模糊数字与现实之间的界线。

关 键 词：MAGIC PROMPT VIDEO 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]