猫咪暴揍哥斯拉、叶问拳打钢铁侠、孙悟空棒打奥特曼、小学生隔扣詹姆斯……最近,Seedance 2.0全球爆火,相关视频也在不少短视频平台刷屏。
Chinese tech company ByteDance has launched the latest version of its artificial intelligence video-generation model, Seedance 2.0, which has quickly gone viral worldwide, stunning creators and prompting some US filmmakers to warn it could upend Hollywood.
upend /ʌpˈend/ 颠覆
近日,字节跳动最新视频生成模型Seedance 2.0在即梦、豆包、小云雀等产品开启内测。
它支持图像、视频、音频、文本四种模态输入,只需输入提示词或上传参考图,就能在60秒内产出带完整原生音轨的多镜头视频,大幅降低专业创作的门槛与成本。
凭借多模态输入、超强参考能力等亮点,Seedance 2.0很快引爆网络。
不仅国内,Seedance 2.0更在海外社交平台上持续刷屏、频频登上热搜。海外用户积极尝试用它生成诸如“悟空对决漫威英雄、宠物猫狗隔扣NBA球星”等天马行空的视频内容,多条演示视频播放量破百万。
Supporting image, video, audio and text inputs, the model can generate multi-shot videos with synchronized soundtracks in about 60 seconds, significantly lowering production costs and technical barriers. Its advanced reference capabilities and multimodal integration have sparked global interest, with imaginative demo clips on overseas social media platforms drawing millions of views.
美国导演:这可能会搞垮好莱坞
据参考消息报道,多位美国导演或影视从业者在使用Seedance 2.0之后,公开表达了诸如“好莱坞要完蛋了”或“这东西说不定真能搞垮好莱坞”的感慨。
Several American directors and film professionals have expressed surprise after testing the tool, with some calling it a potential disruptor of traditional filmmaking.
导演Charles Curran在社交平台连续发文,他为从未面世的真人电影制作了一版预告片,仅用20分钟,花费60美元,借助Seedance 2.0完成,“这说不定真能搞砸好莱坞呢”。
在他发布的1分24秒的视频里,做到了电影级画面特效、音画同步、多机位“拍摄”等。
US documentary filmmaker Charles Curran said he used the model to produce a trailer for an unreleased live-action film in 20 minutes at a cost of $60. The 1-minute-24-second clip featured cinematic visual effects, synchronized audio and simulated multi-camera shooting.
据了解,Charles Curran是美国纽约资深纪录片导演、制片人、剪辑师,长期深耕纪实影像与实验电影创作,同时为国际顶级品牌提供商业影像服务。
不少科技工作者和AI研究者也表达了类似的观点:
另据南都周刊报道,对于网友提出的可能混淆现实的担忧,2月9日,即梦创作者社群中,平台运营人员发布消息称:“Seedance 2.0在内测期间收获了远超预期的关注,感谢大家的使用反馈。为了保障创作环境的健康可持续,我们正在针对反馈进行紧急优化,目前暂不支持输入真人图片或视频作为主体参考”,并表示平台深知创意的边界是尊重,产品调整后会以更完善的面貌与大家正式见面。
Addressing concerns about AI-generated content blurring the line with reality, the platform said it is making urgent adjustments following beta feedback. It will temporarily disable the use of real-person images or videos as reference material while refining the product.
来源:中国青年报综合南都周刊、参考消息、蓝鲸新闻、21世纪经济报道等
跟着China Daily
精读英语新闻
“无痛”学英语,每天20分钟就够!